Wiki - Факультет компьютерных наук - Вклад [ru]

Глубинное обучение 2 2025

2025-12-20T15:49:48Z

Anapluslap: add materials for class 13

== Общая информация ==
Курс предназначен для студентов 4 курса ФКН ПМИ (МОП и КНАД).

Занятия проходят '''по понедельникам 14:40-17:40''' (переносы будут сообщаться в чате).

Полезные ссылки:
* Чат с обсуждением: https://t.me/+8dcjl4gHlyEwYzcy
* Репозиторий курса: https://github.com/thecrazymage/DL2_HSE
* Таблица с оценками: https://docs.google.com/spreadsheets/d/18rfZbf7Zsm-xmbiQn_eJVDXXbDfesbHT/edit?usp=sharing&ouid=115132401804687564737&rtpof=true&sd=true
* Anytask: https://anytask.org/course/1219

Формула итоговой оценки (округление арифметическое):
# МОП: Оитог = 0.25 * Осоревнование + 0.75 * ОДЗ,
# КНАД: Оитог = ОДЗ,
где ОДЗ - средняя оценка за практические домашние задания.

== Преподаватели и ассистенты ==

Кому писать, если кажется, что все пропало: [https://t.me/MishanAliev Мишан Алиев]

{| class="wikitable"
|-
! Группа !! Семинарист !! Ассистенты !! Чаты групп !! Инвайт в anytask
|-
| 221 (МОП) || [https://t.me/mightyneighbor Федя Великонивцев] || [https://t.me/vsem_paket Динар Саберов], [https://t.me/annvasileeva Анна Василева] || [https://t.me/+XzlFh0Y8GHs3NTMy МОП 221] || yvZZTIS
|-
| 222 (МОП) || [https://t.me/cocosinca Ева Неудачина] || [https://t.me/mat_os Александр Матосян], [https://t.me/kade1shvili Полина Кадейшвили] || [https://t.me/+PpxiIpF7wLYyMjE6 МОП 222] || W38CfZf
|-
| 223 (МОП) || [https://t.me/tutugarin Иван Ершов] || [https://t.me/Andrew_ut Андрей Уткин], [https://t.me/uzhedevyat Георгий Фатахов] || [https://t.me/+xRZkDO4TjiA1YjQ6 МОП 223] || ph91Jlz
|-
| 224 (МОП) || [https://t.me/Stepyndriy Степан Беляков] || [https://t.me/ponomarchuk_anna Анна Пономарчук], [https://t.me/bogsolntca Татьяна Яковлева] || [https://t.me/+XtcOG352A-IxMmY6 МОП 224] || iPwd342
|-
| КНАД || [https://t.me/burakevi4 Даня Бураков] || [https://t.me/anapluslap Анастасия Лапшина], [https://t.me/irbix7 Иван Галий] || [https://t.me/+JNPUgVxh6ophOGQy КНАД] || zTn4sRP
|}

== Лекции и семинары ==

'''Лекция / Семинар 1 (08.09).'''

Тема: Essentials of GPU, Deep Learning Bottlenecks, and Benchmarking Basics

Аннотация: In this session, we will explore the reasons behind the dominance of GPUs in Deep Learning and examine the common sources of performance bottlenecks in DL code. You will learn how to identify these bottlenecks using profiling tools and apply techniques to optimize and accelerate your code.

Лектор и семинарист: [https://www.hse.ru/org/persons/816100677/ Fedor Velikonivtsev]

Локация: лекция - R208, семинар - D102.

Материалы: [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/lecture_01.mp4 запись лекции], [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/seminar_01.mp4 запись семинара], [https://github.com/thecrazymage/DL2_HSE/tree/main/week_01 материалы].

'''Лекция / Семинар 2 (15.09).'''

Тема: On Transformers and Bitter Lesson

Аннотация: In this talk we’ll dive into the landscape of model architectures in deep learning, with a focus on the world around transformers. We’ll briefly recall what a transformer is, trace the evolution from encoder–decoder to encoder-only and decoder-only models, and touch on the rise of “efficient mixers” such as state space models, linear attention, and beyond. We’ll conclude by reflecting on the role of data and compute. The lecture is inspired in part by [http://incompleteideas.net/IncIdeas/BitterLesson.html The Bitter Lesson] and blends a brain dump with some entertaining insights from recent years of architectural exploration.

Лектор и семинарист: [https://www.hse.ru/org/persons/190912012/ Ivan Rubachev]

Локация: лекция - R206, семинар - R208.

Материалы: [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/lecture%2Bseminar_02.mp4 запись лекции и семинара], [https://github.com/thecrazymage/DL2_HSE/tree/main/week_02 материалы].

'''Лекция / Семинар 3 (22.09).'''

Тема: Modern LLMs essentials

Аннотация лекции: This week, we will discuss LLMs. We will discuss why they are so effective for text generation, how they can be applied to different NLP problems, and the risks they pose. You will learn the details of RLHF, PEFT, and RAG, which make LLMs robust in various cases.

Лектор: [https://www.hse.ru/org/persons/208533329/ Alexander Shabalin]

Локация: онлайн.

Аннотация семинара: In this seminar, we will explore the concept of LLM-based agents and how they extend the capabilities of modern language models. We will discuss function calling as a way to integrate external tools, chain-of-thought reasoning for structured problem solving, and reinforcement learning techniques for training agents.

Семинарист: [http://www.instagram.com/tugarin_vanya Ivan Ershov]

Локация: R208.

Материалы: [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/lecture_03.mp4 запись лекции], [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/seminar_03.mp4 запись семинара], [https://github.com/thecrazymage/DL2_HSE/tree/main/week_03 материалы].

'''Лекция / Семинар 4 (29.09).'''

Тема: Basics of Efficient LLM Training Infrastructure

Аннотация: In this lecture, we will study the basic rules that underpin the infrastructure for efficient large language model training. We will also examine common problems that may arise in this process and explore practical ways to address them.

Лектор и семинарист: [https://habr.com/ru/users/MichaelEk/ Michael Khrushchev]

Локация: лекция - G002, семинар - R208.

Материалы: [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/lecture%2Bseminar_04.mp4 запись лекции и семинара], [https://github.com/thecrazymage/DL2_HSE/tree/main/week_04 материалы].

'''Лекция / Семинар 5 (06.10).'''

Тема: Segmentation and Detection

Аннотация лекции: This week, we'll explore the evolution of object detection and segmentation — from R-CNN to real-time methods like YOLO-World and CLIP-based approaches. We'll examine how U-Net architectures have transcended computer vision to power neural operators and how to use diffusion models for segmentation tasks.

Аннотация семинара: We will implement architectures and train a semantic segmentation model. We will discuss regularization methods for convolutional layers. In addition, we will learn how to use pre-trained models and apply them.

Лектор и семинарист: [https://www.hse.ru/org/persons/956617478/ Alexander Oganov]

Локация: лекция - G002, семинар - R208.

Материалы: [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/lecture_05.mp4 запись лекции], [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/seminar_05.mp4 запись семинара], [https://github.com/thecrazymage/DL2_HSE/tree/main/week_05 материалы].

'''Лекция / Семинар 6 (13.10).'''

Тема: Segmentation and Detection 2

Аннотация лекции и семинара: In this lecture, we will continue in more detail about segmentation and detection.

Лектор: [https://szagoruyko.github.io/ Sergey Zagoruyko]

Семинарист: [https://www.hse.ru/org/persons/401628708/ Eva Neudachina]

Локация: лекция - G002, семинар - R208.

Материалы: [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/lecture_06.mp4 запись лекции], [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/seminar_06.mp4 запись семинара], [https://github.com/thecrazymage/DL2_HSE/tree/main/week_06 материалы].

'''Лекция / Семинар 7 (20.10).'''

Тема: Diffusion models 1

Аннотация лекции и семинара: Tomorrow's lecture and seminar will be devoted to an introduction to diffusion models. Diffusion models are currently the most popular approach to generative modeling due to their high-quality generation and diversity (mode coverage) of the learned distribution. The idea behind diffusion models is to consider the process of gradually transforming data into pure noise and construct its inverse in time, which will transform noise into data. In the lecture and seminar, we will work with noise processes and derive the classic [https://arxiv.org/abs/2006.11239 DDPM] model, which proposes to minimize the KL-divergence between the “true” reverse process that converts noise into data and the denoising process specified by the neural network. In the process, we will see that this procedure is equivalent to training a denoiser neural network that predicts a clean object from a noisy one. In addition, we will interpret the resulting denoising process: in it, each step corresponds to replacing part of the current noisy image with an (increasingly high-quality) prediction of the denoiser.

Лектор и семинарист: [https://www.hse.ru/org/persons/190910999/ Denis Rakitin]

Локация: лекция - G002, семинар - R208.

Материалы: [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/lecture_07.mp4 запись лекции], [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/seminar_07.mp4 запись семинара], [https://github.com/thecrazymage/DL2_HSE/tree/main/week_07 материалы].

'''Лекция / Семинар 8 (03.11).'''

Тема: Diffusion models 2

Аннотация лекции и семинара: In this lecture and seminar, we will continue our exploration of diffusion models. We will introduce the score function and score identity, present classifier and classifier-free guidance, and derive DDIM model.

Лектор и семинарист: [https://www.hse.ru/org/persons/190910999/ Denis Rakitin]

Локация: лекция - R205, семинар - R208.

Материалы: [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/lecture_08.mp4 запись лекции], [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/seminar_08.mp4 запись семинара], [https://github.com/thecrazymage/DL2_HSE/tree/main/week_08 материалы].

'''Лекция / Семинар 9 (10.11).'''

Тема: Diffusion models 3

Аннотация лекции: This lecture presents applied aspects of diffusion models, a class of generative methods that have demonstrated strong performance across images, text, and audio. We will review modern architectures, with a focus on [https://github.com/black-forest-labs/flux FLUX], and unpack the key principles of their design and training.

Лектор: [https://scholar.google.com/citations?user=o6pRm_gAAAAJ&hl=ru Nikita Starodubcev]

Аннотация семинара: The seminar will provide a hands-on, in-depth inspection of [https://github.com/PixArt-alpha/PixArt-alpha PixArt] model. Then ee will attempt further fine-tuning via [https://dreambooth.github.io/ DreamBooth], review common evaluation metrics, and compare different models and configurations.

Семинарист: [https://www.hse.ru/org/persons/401628708/ Eva Neudachina]

Локация: лекция - R205, семинар - R208.

Материалы: [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/lecture_09.mp4 запись лекции], [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/seminar_09.mp4 запись семинара], [https://github.com/thecrazymage/DL2_HSE/tree/main/week_09 материалы].

'''Лекция / Семинар 10 (17.11).'''

Тема: 3D CV

Аннотация лекции и семинара: This lecture and seminar will discuss how diffusion models can be used for 3D computer vision tasks such as generation and reconstruction, and how they relate to methods like NeRF and Gaussian splatting. We will also look at how these models can be applied to problems like novel view synthesis, relighting, and camera pose or depth estimation.

Лектор и семинарист: [https://www.hse.ru/org/persons/885876805/ Mishan Aliev]

Локация: лекция - R205, семинар - R208.

Материалы: [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/lecture_10.mp4 запись лекции], [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/seminar_10.mp4 запись семинара], [https://github.com/thecrazymage/DL2_HSE/tree/main/week_10 материалы].

'''Лекция / Семинар 11 (24.11).'''

Тема: Neural Recommender Systems

Аннотация лекции и семинара: Modern recommender systems also make heavy use of deep learning — both discriminative and generative models. A system has to understand content (tracks, products, videos, etc.), model user behavior (short- and long-term preferences), and predict which content each user is likely to enjoy. At the same time, recommender systems come with their own challenges: cold start for new items and users, huge billion-scale item catalogs with heavy-tailed popularity distributions (popularity bias), and constant distribution drift, where all the underlying distributions keep changing along with the rest of the world. We will discuss: 1) Two-tower neural networks for candidate generation; 2) Sequential recommendation and more modern approaches to generative modeling of users; 3) Semantic IDs and tuning LLMs for recommendation 4) What makes the recommendation domain different from other deep learning domains.

Лектор: [https://www.hse.ru/org/persons/1085421573/ Kirill Khrylchenko]

Семинарист: [https://www.linkedin.com/in/artem-matveev-7b2725255/ Artem Matveev]

Локация: лекция - R205, семинар - R208.

Материалы: [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/lecture_11.mp4 запись лекции], [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/seminar_11.mp4 запись семинара], [https://github.com/thecrazymage/DL2_HSE/tree/main/week_11 материалы].

'''Лекция / Семинар 12 (08.12).'''

Тема: Graph Machine Learning

Аннотация лекции и семинара: In the lecture, we will explore the basics of Graph Machine Learning (Graph-ML), the tasks are being solved in this domain, and the properties of the models and architectures being used in Graph-ML. In the seminar, we will train our Graph Neural Network using modern Graph-ML frameworks.

Лектор и семинарист: [https://www.hse.ru/org/persons/816100677/ Fedor Velikonivtsev]

Локация: лекция - R205, семинар - R208.

Материалы: [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/lecture_12.mp4 запись лекции], [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/seminar_12.mp4 запись семинара], [https://github.com/thecrazymage/DL2_HSE/tree/main/week_12 материалы].

'''Лекция / Семинар 13 (15.12).'''

Тема: Deep Learning and Tabular Data

Аннотация лекции: We'll talk about DL and tabular data. We'll go from [https://scikit-learn.org/stable/modules/generated/sklearn.datasets.fetch_california_housing.html California] to $10М [https://www.eu-startups.com/2025/02/prior-labs-raises-e9-million-for-foundation-models-for-spreadsheets-and-databases/ seed rounds]. I plan to give a broad outlook on the field, trace how it evolved and talk a bit about where it's heading. We'll talk about DL model components specific to tabular data, some speculations and intuitions on why those methods may work, and finish off with a talk about foundation models for tabular data.

Аннотация семинара: Hands-on tabular DL — on a MacBook. The idea: implement some of the core models from scratch (or watch me do it). The hardware constraint might help us appreciate the practical limits (or lack thereof).

Лектор и семинарист: [https://www.hse.ru/org/persons/190912012/ Ivan Rubachev]

Локация: лекция - R205, семинар - R208.

Материалы: [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/lecture_13.mp4 запись лекции], [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/seminar_13.mp4 запись семинара], [https://github.com/thecrazymage/DL2_HSE/tree/main/week_13 материалы].

== Домашние задания ==

{| class="wikitable"
! № !! Домашнее задание !! Ссылка !! Дедлайн (жёсткий)
|-
| 1 || Tensor and DL Libraries || https://github.com/thecrazymage/DL2_HSE/tree/main/homeworks/homework_01 || 30 сентября, 23:59
|-
| 2 || Transformers for Named Entity Recognition || https://github.com/thecrazymage/DL2_HSE/tree/main/homeworks/homework_02 || 14 октября, 23:59
|-
| 3 || Image Segmentation || https://github.com/thecrazymage/DL2_HSE/tree/main/homeworks/homework_03 || 2 ноября, 23:59
|-
| 4 || Diffusion Models || https://github.com/thecrazymage/DL2_HSE/tree/main/homeworks/homework_04 || 18 ноября, 23:59
|-
| 5 || Retrieval || https://github.com/thecrazymage/DL2_HSE/tree/main/homeworks/homework_05 || 17 декабря, 23:59

|}

== Соревнование ==

'''Задание:''' https://github.com/thecrazymage/DL2_HSE/tree/main/homeworks/competition

'''Возможность перезачета:''' https://t.me/c/2969026465/2/1601.

'''Дедлайн (жёсткий):''' 19 декабря 2025, 23:59

Глубинное обучение 2 2025

2025-12-15T11:43:08Z

Anapluslap: add class 13 announc. and materials for 12

== Общая информация ==
Курс предназначен для студентов 4 курса ФКН ПМИ (МОП и КНАД).

Занятия проходят '''по понедельникам 14:40-17:40''' (переносы будут сообщаться в чате).

Полезные ссылки:
* Чат с обсуждением: https://t.me/+8dcjl4gHlyEwYzcy
* Репозиторий курса: https://github.com/thecrazymage/DL2_HSE
* Таблица с оценками: https://docs.google.com/spreadsheets/d/18rfZbf7Zsm-xmbiQn_eJVDXXbDfesbHT/edit?usp=sharing&ouid=115132401804687564737&rtpof=true&sd=true
* Anytask: https://anytask.org/course/1219

Формула итоговой оценки (округление арифметическое):
# МОП: Оитог = 0.25 * Осоревнование + 0.75 * ОДЗ,
# КНАД: Оитог = ОДЗ,
где ОДЗ - средняя оценка за практические домашние задания.

== Преподаватели и ассистенты ==

Кому писать, если кажется, что все пропало: [https://t.me/MishanAliev Мишан Алиев]

{| class="wikitable"
|-
! Группа !! Семинарист !! Ассистенты !! Чаты групп !! Инвайт в anytask
|-
| 221 (МОП) || [https://t.me/mightyneighbor Федя Великонивцев] || [https://t.me/vsem_paket Динар Саберов], [https://t.me/annvasileeva Анна Василева] || [https://t.me/+XzlFh0Y8GHs3NTMy МОП 221] || yvZZTIS
|-
| 222 (МОП) || [https://t.me/cocosinca Ева Неудачина] || [https://t.me/mat_os Александр Матосян], [https://t.me/kade1shvili Полина Кадейшвили] || [https://t.me/+PpxiIpF7wLYyMjE6 МОП 222] || W38CfZf
|-
| 223 (МОП) || [https://t.me/tutugarin Иван Ершов] || [https://t.me/Andrew_ut Андрей Уткин], [https://t.me/uzhedevyat Георгий Фатахов] || [https://t.me/+xRZkDO4TjiA1YjQ6 МОП 223] || ph91Jlz
|-
| 224 (МОП) || [https://t.me/Stepyndriy Степан Беляков] || [https://t.me/ponomarchuk_anna Анна Пономарчук], [https://t.me/bogsolntca Татьяна Яковлева] || [https://t.me/+XtcOG352A-IxMmY6 МОП 224] || iPwd342
|-
| КНАД || [https://t.me/burakevi4 Даня Бураков] || [https://t.me/anapluslap Анастасия Лапшина], [https://t.me/irbix7 Иван Галий] || [https://t.me/+JNPUgVxh6ophOGQy КНАД] || zTn4sRP
|}

== Лекции и семинары ==

'''Лекция / Семинар 1 (08.09).'''

Тема: Essentials of GPU, Deep Learning Bottlenecks, and Benchmarking Basics

Аннотация: In this session, we will explore the reasons behind the dominance of GPUs in Deep Learning and examine the common sources of performance bottlenecks in DL code. You will learn how to identify these bottlenecks using profiling tools and apply techniques to optimize and accelerate your code.

Лектор и семинарист: [https://www.hse.ru/org/persons/816100677/ Fedor Velikonivtsev]

Локация: лекция - R208, семинар - D102.

Материалы: [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/lecture_01.mp4 запись лекции], [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/seminar_01.mp4 запись семинара], [https://github.com/thecrazymage/DL2_HSE/tree/main/week_01 материалы].

'''Лекция / Семинар 2 (15.09).'''

Тема: On Transformers and Bitter Lesson

Аннотация: In this talk we’ll dive into the landscape of model architectures in deep learning, with a focus on the world around transformers. We’ll briefly recall what a transformer is, trace the evolution from encoder–decoder to encoder-only and decoder-only models, and touch on the rise of “efficient mixers” such as state space models, linear attention, and beyond. We’ll conclude by reflecting on the role of data and compute. The lecture is inspired in part by [http://incompleteideas.net/IncIdeas/BitterLesson.html The Bitter Lesson] and blends a brain dump with some entertaining insights from recent years of architectural exploration.

Лектор и семинарист: [https://www.hse.ru/org/persons/190912012/ Ivan Rubachev]

Локация: лекция - R206, семинар - R208.

Материалы: [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/lecture%2Bseminar_02.mp4 запись лекции и семинара], [https://github.com/thecrazymage/DL2_HSE/tree/main/week_02 материалы].

'''Лекция / Семинар 3 (22.09).'''

Тема: Modern LLMs essentials

Аннотация лекции: This week, we will discuss LLMs. We will discuss why they are so effective for text generation, how they can be applied to different NLP problems, and the risks they pose. You will learn the details of RLHF, PEFT, and RAG, which make LLMs robust in various cases.

Лектор: [https://www.hse.ru/org/persons/208533329/ Alexander Shabalin]

Локация: онлайн.

Аннотация семинара: In this seminar, we will explore the concept of LLM-based agents and how they extend the capabilities of modern language models. We will discuss function calling as a way to integrate external tools, chain-of-thought reasoning for structured problem solving, and reinforcement learning techniques for training agents.

Семинарист: [http://www.instagram.com/tugarin_vanya Ivan Ershov]

Локация: R208.

Материалы: [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/lecture_03.mp4 запись лекции], [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/seminar_03.mp4 запись семинара], [https://github.com/thecrazymage/DL2_HSE/tree/main/week_03 материалы].

'''Лекция / Семинар 4 (29.09).'''

Тема: Basics of Efficient LLM Training Infrastructure

Аннотация: In this lecture, we will study the basic rules that underpin the infrastructure for efficient large language model training. We will also examine common problems that may arise in this process and explore practical ways to address them.

Лектор и семинарист: [https://habr.com/ru/users/MichaelEk/ Michael Khrushchev]

Локация: лекция - G002, семинар - R208.

Материалы: [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/lecture%2Bseminar_04.mp4 запись лекции и семинара], [https://github.com/thecrazymage/DL2_HSE/tree/main/week_04 материалы].

'''Лекция / Семинар 5 (06.10).'''

Тема: Segmentation and Detection

Аннотация лекции: This week, we'll explore the evolution of object detection and segmentation — from R-CNN to real-time methods like YOLO-World and CLIP-based approaches. We'll examine how U-Net architectures have transcended computer vision to power neural operators and how to use diffusion models for segmentation tasks.

Аннотация семинара: We will implement architectures and train a semantic segmentation model. We will discuss regularization methods for convolutional layers. In addition, we will learn how to use pre-trained models and apply them.

Лектор и семинарист: [https://www.hse.ru/org/persons/956617478/ Alexander Oganov]

Локация: лекция - G002, семинар - R208.

Материалы: [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/lecture_05.mp4 запись лекции], [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/seminar_05.mp4 запись семинара], [https://github.com/thecrazymage/DL2_HSE/tree/main/week_05 материалы].

'''Лекция / Семинар 6 (13.10).'''

Тема: Segmentation and Detection 2

Аннотация лекции и семинара: In this lecture, we will continue in more detail about segmentation and detection.

Лектор: [https://szagoruyko.github.io/ Sergey Zagoruyko]

Семинарист: [https://www.hse.ru/org/persons/401628708/ Eva Neudachina]

Локация: лекция - G002, семинар - R208.

Материалы: [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/lecture_06.mp4 запись лекции], [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/seminar_06.mp4 запись семинара], [https://github.com/thecrazymage/DL2_HSE/tree/main/week_06 материалы].

'''Лекция / Семинар 7 (20.10).'''

Тема: Diffusion models 1

Аннотация лекции и семинара: Tomorrow's lecture and seminar will be devoted to an introduction to diffusion models. Diffusion models are currently the most popular approach to generative modeling due to their high-quality generation and diversity (mode coverage) of the learned distribution. The idea behind diffusion models is to consider the process of gradually transforming data into pure noise and construct its inverse in time, which will transform noise into data. In the lecture and seminar, we will work with noise processes and derive the classic [https://arxiv.org/abs/2006.11239 DDPM] model, which proposes to minimize the KL-divergence between the “true” reverse process that converts noise into data and the denoising process specified by the neural network. In the process, we will see that this procedure is equivalent to training a denoiser neural network that predicts a clean object from a noisy one. In addition, we will interpret the resulting denoising process: in it, each step corresponds to replacing part of the current noisy image with an (increasingly high-quality) prediction of the denoiser.

Лектор и семинарист: [https://www.hse.ru/org/persons/190910999/ Denis Rakitin]

Локация: лекция - G002, семинар - R208.

Материалы: [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/lecture_07.mp4 запись лекции], [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/seminar_07.mp4 запись семинара], [https://github.com/thecrazymage/DL2_HSE/tree/main/week_07 материалы].

'''Лекция / Семинар 8 (03.11).'''

Тема: Diffusion models 2

Аннотация лекции и семинара: In this lecture and seminar, we will continue our exploration of diffusion models. We will introduce the score function and score identity, present classifier and classifier-free guidance, and derive DDIM model.

Лектор и семинарист: [https://www.hse.ru/org/persons/190910999/ Denis Rakitin]

Локация: лекция - R205, семинар - R208.

Материалы: [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/lecture_08.mp4 запись лекции], [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/seminar_08.mp4 запись семинара], [https://github.com/thecrazymage/DL2_HSE/tree/main/week_08 материалы].

'''Лекция / Семинар 9 (10.11).'''

Тема: Diffusion models 3

Аннотация лекции: This lecture presents applied aspects of diffusion models, a class of generative methods that have demonstrated strong performance across images, text, and audio. We will review modern architectures, with a focus on [https://github.com/black-forest-labs/flux FLUX], and unpack the key principles of their design and training.

Лектор: [https://scholar.google.com/citations?user=o6pRm_gAAAAJ&hl=ru Nikita Starodubcev]

Аннотация семинара: The seminar will provide a hands-on, in-depth inspection of [https://github.com/PixArt-alpha/PixArt-alpha PixArt] model. Then ee will attempt further fine-tuning via [https://dreambooth.github.io/ DreamBooth], review common evaluation metrics, and compare different models and configurations.

Семинарист: [https://www.hse.ru/org/persons/401628708/ Eva Neudachina]

Локация: лекция - R205, семинар - R208.

Материалы: [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/lecture_09.mp4 запись лекции], [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/seminar_09.mp4 запись семинара], [https://github.com/thecrazymage/DL2_HSE/tree/main/week_09 материалы].

'''Лекция / Семинар 10 (17.11).'''

Тема: 3D CV

Аннотация лекции и семинара: This lecture and seminar will discuss how diffusion models can be used for 3D computer vision tasks such as generation and reconstruction, and how they relate to methods like NeRF and Gaussian splatting. We will also look at how these models can be applied to problems like novel view synthesis, relighting, and camera pose or depth estimation.

Лектор и семинарист: [https://www.hse.ru/org/persons/885876805/ Mishan Aliev]

Локация: лекция - R205, семинар - R208.

Материалы: [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/lecture_10.mp4 запись лекции], [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/seminar_10.mp4 запись семинара], [https://github.com/thecrazymage/DL2_HSE/tree/main/week_10 материалы].

'''Лекция / Семинар 11 (24.11).'''

Тема: Neural Recommender Systems

Аннотация лекции и семинара: Modern recommender systems also make heavy use of deep learning — both discriminative and generative models. A system has to understand content (tracks, products, videos, etc.), model user behavior (short- and long-term preferences), and predict which content each user is likely to enjoy. At the same time, recommender systems come with their own challenges: cold start for new items and users, huge billion-scale item catalogs with heavy-tailed popularity distributions (popularity bias), and constant distribution drift, where all the underlying distributions keep changing along with the rest of the world. We will discuss: 1) Two-tower neural networks for candidate generation; 2) Sequential recommendation and more modern approaches to generative modeling of users; 3) Semantic IDs and tuning LLMs for recommendation 4) What makes the recommendation domain different from other deep learning domains.

Лектор: [https://www.hse.ru/org/persons/1085421573/ Kirill Khrylchenko]

Семинарист: [https://www.linkedin.com/in/artem-matveev-7b2725255/ Artem Matveev]

Локация: лекция - R205, семинар - R208.

Материалы: [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/lecture_11.mp4 запись лекции], [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/seminar_11.mp4 запись семинара], [https://github.com/thecrazymage/DL2_HSE/tree/main/week_11 материалы].

'''Лекция / Семинар 12 (08.12).'''

Тема: Graph Machine Learning

Аннотация лекции и семинара: In the lecture, we will explore the basics of Graph Machine Learning (Graph-ML), the tasks are being solved in this domain, and the properties of the models and architectures being used in Graph-ML. In the seminar, we will train our Graph Neural Network using modern Graph-ML frameworks.

Лектор и семинарист: [https://www.hse.ru/org/persons/816100677/ Fedor Velikonivtsev]

Локация: лекция - R205, семинар - R208.

Материалы: [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/lecture_12.mp4 запись лекции], [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/seminar_12.mp4 запись семинара], [https://github.com/thecrazymage/DL2_HSE/tree/main/week_12 материалы].

'''Лекция / Семинар 13 (15.12).'''

Тема: Deep Learning and Tabular Data

Аннотация лекции: We'll talk about DL and tabular data. We'll go from [https://scikit-learn.org/stable/modules/generated/sklearn.datasets.fetch_california_housing.html California] to $10М [https://www.eu-startups.com/2025/02/prior-labs-raises-e9-million-for-foundation-models-for-spreadsheets-and-databases/ seed rounds]. I plan to give a broad outlook on the field, trace how it evolved and talk a bit about where it's heading. We'll talk about DL model components specific to tabular data, some speculations and intuitions on why those methods may work, and finish off with a talk about foundation models for tabular data.

Аннотация семинара: Hands-on tabular DL — on a MacBook. The idea: implement some of the core models from scratch (or watch me do it). The hardware constraint might help us appreciate the practical limits (or lack thereof).

Лектор и семинарист: [https://www.hse.ru/org/persons/190912012/ Ivan Rubachev]

Локация: лекция - R205, семинар - R208.

Материалы: tba.

== Домашние задания ==

{| class="wikitable"
! № !! Домашнее задание !! Ссылка !! Дедлайн (жёсткий)
|-
| 1 || Tensor and DL Libraries || https://github.com/thecrazymage/DL2_HSE/tree/main/homeworks/homework_01 || 30 сентября, 23:59
|-
| 2 || Transformers for Named Entity Recognition || https://github.com/thecrazymage/DL2_HSE/tree/main/homeworks/homework_02 || 14 октября, 23:59
|-
| 3 || Image Segmentation || https://github.com/thecrazymage/DL2_HSE/tree/main/homeworks/homework_03 || 2 ноября, 23:59
|-
| 4 || Diffusion Models || https://github.com/thecrazymage/DL2_HSE/tree/main/homeworks/homework_04 || 18 ноября, 23:59
|-
| 5 || Retrieval || https://github.com/thecrazymage/DL2_HSE/tree/main/homeworks/homework_05 || 17 декабря, 23:59

|}

== Соревнование ==

'''Задание:''' https://github.com/thecrazymage/DL2_HSE/tree/main/homeworks/competition

'''Возможность перезачета:''' https://t.me/c/2969026465/2/1601.

'''Дедлайн (жёсткий):''' 19 декабря 2025, 23:59

Глубинное обучение 2 2025

2025-12-12T10:09:12Z

Anapluslap: change deadline for hw5

== Общая информация ==
Курс предназначен для студентов 4 курса ФКН ПМИ (МОП и КНАД).

Занятия проходят '''по понедельникам 14:40-17:40''' (переносы будут сообщаться в чате).

Полезные ссылки:
* Чат с обсуждением: https://t.me/+8dcjl4gHlyEwYzcy
* Репозиторий курса: https://github.com/thecrazymage/DL2_HSE
* Таблица с оценками: https://docs.google.com/spreadsheets/d/18rfZbf7Zsm-xmbiQn_eJVDXXbDfesbHT/edit?usp=sharing&ouid=115132401804687564737&rtpof=true&sd=true
* Anytask: https://anytask.org/course/1219

Формула итоговой оценки (округление арифметическое):
# МОП: Оитог = 0.25 * Осоревнование + 0.75 * ОДЗ,
# КНАД: Оитог = ОДЗ,
где ОДЗ - средняя оценка за практические домашние задания.

== Преподаватели и ассистенты ==

Кому писать, если кажется, что все пропало: [https://t.me/MishanAliev Мишан Алиев]

{| class="wikitable"
|-
! Группа !! Семинарист !! Ассистенты !! Чаты групп !! Инвайт в anytask
|-
| 221 (МОП) || [https://t.me/mightyneighbor Федя Великонивцев] || [https://t.me/vsem_paket Динар Саберов], [https://t.me/annvasileeva Анна Василева] || [https://t.me/+XzlFh0Y8GHs3NTMy МОП 221] || yvZZTIS
|-
| 222 (МОП) || [https://t.me/cocosinca Ева Неудачина] || [https://t.me/mat_os Александр Матосян], [https://t.me/kade1shvili Полина Кадейшвили] || [https://t.me/+PpxiIpF7wLYyMjE6 МОП 222] || W38CfZf
|-
| 223 (МОП) || [https://t.me/tutugarin Иван Ершов] || [https://t.me/Andrew_ut Андрей Уткин], [https://t.me/uzhedevyat Георгий Фатахов] || [https://t.me/+xRZkDO4TjiA1YjQ6 МОП 223] || ph91Jlz
|-
| 224 (МОП) || [https://t.me/Stepyndriy Степан Беляков] || [https://t.me/ponomarchuk_anna Анна Пономарчук], [https://t.me/bogsolntca Татьяна Яковлева] || [https://t.me/+XtcOG352A-IxMmY6 МОП 224] || iPwd342
|-
| КНАД || [https://t.me/burakevi4 Даня Бураков] || [https://t.me/anapluslap Анастасия Лапшина], [https://t.me/irbix7 Иван Галий] || [https://t.me/+JNPUgVxh6ophOGQy КНАД] || zTn4sRP
|}

== Лекции и семинары ==

'''Лекция / Семинар 1 (08.09).'''

Тема: Essentials of GPU, Deep Learning Bottlenecks, and Benchmarking Basics

Аннотация: In this session, we will explore the reasons behind the dominance of GPUs in Deep Learning and examine the common sources of performance bottlenecks in DL code. You will learn how to identify these bottlenecks using profiling tools and apply techniques to optimize and accelerate your code.

Лектор и семинарист: [https://www.hse.ru/org/persons/816100677/ Fedor Velikonivtsev]

Локация: лекция - R208, семинар - D102.

Материалы: [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/lecture_01.mp4 запись лекции], [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/seminar_01.mp4 запись семинара], [https://github.com/thecrazymage/DL2_HSE/tree/main/week_01 материалы].

'''Лекция / Семинар 2 (15.09).'''

Тема: On Transformers and Bitter Lesson

Аннотация: In this talk we’ll dive into the landscape of model architectures in deep learning, with a focus on the world around transformers. We’ll briefly recall what a transformer is, trace the evolution from encoder–decoder to encoder-only and decoder-only models, and touch on the rise of “efficient mixers” such as state space models, linear attention, and beyond. We’ll conclude by reflecting on the role of data and compute. The lecture is inspired in part by [http://incompleteideas.net/IncIdeas/BitterLesson.html The Bitter Lesson] and blends a brain dump with some entertaining insights from recent years of architectural exploration.

Лектор и семинарист: [https://www.hse.ru/org/persons/190912012/ Ivan Rubachev]

Локация: лекция - R206, семинар - R208.

Материалы: [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/lecture%2Bseminar_02.mp4 запись лекции и семинара], [https://github.com/thecrazymage/DL2_HSE/tree/main/week_02 материалы].

'''Лекция / Семинар 3 (22.09).'''

Тема: Modern LLMs essentials

Аннотация лекции: This week, we will discuss LLMs. We will discuss why they are so effective for text generation, how they can be applied to different NLP problems, and the risks they pose. You will learn the details of RLHF, PEFT, and RAG, which make LLMs robust in various cases.

Лектор: [https://www.hse.ru/org/persons/208533329/ Alexander Shabalin]

Локация: онлайн.

Аннотация семинара: In this seminar, we will explore the concept of LLM-based agents and how they extend the capabilities of modern language models. We will discuss function calling as a way to integrate external tools, chain-of-thought reasoning for structured problem solving, and reinforcement learning techniques for training agents.

Семинарист: [http://www.instagram.com/tugarin_vanya Ivan Ershov]

Локация: R208.

Материалы: [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/lecture_03.mp4 запись лекции], [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/seminar_03.mp4 запись семинара], [https://github.com/thecrazymage/DL2_HSE/tree/main/week_03 материалы].

'''Лекция / Семинар 4 (29.09).'''

Тема: Basics of Efficient LLM Training Infrastructure

Аннотация: In this lecture, we will study the basic rules that underpin the infrastructure for efficient large language model training. We will also examine common problems that may arise in this process and explore practical ways to address them.

Лектор и семинарист: [https://habr.com/ru/users/MichaelEk/ Michael Khrushchev]

Локация: лекция - G002, семинар - R208.

Материалы: [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/lecture%2Bseminar_04.mp4 запись лекции и семинара], [https://github.com/thecrazymage/DL2_HSE/tree/main/week_04 материалы].

'''Лекция / Семинар 5 (06.10).'''

Тема: Segmentation and Detection

Аннотация лекции: This week, we'll explore the evolution of object detection and segmentation — from R-CNN to real-time methods like YOLO-World and CLIP-based approaches. We'll examine how U-Net architectures have transcended computer vision to power neural operators and how to use diffusion models for segmentation tasks.

Аннотация семинара: We will implement architectures and train a semantic segmentation model. We will discuss regularization methods for convolutional layers. In addition, we will learn how to use pre-trained models and apply them.

Лектор и семинарист: [https://www.hse.ru/org/persons/956617478/ Alexander Oganov]

Локация: лекция - G002, семинар - R208.

Материалы: [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/lecture_05.mp4 запись лекции], [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/seminar_05.mp4 запись семинара], [https://github.com/thecrazymage/DL2_HSE/tree/main/week_05 материалы].

'''Лекция / Семинар 6 (13.10).'''

Тема: Segmentation and Detection 2

Аннотация лекции и семинара: In this lecture, we will continue in more detail about segmentation and detection.

Лектор: [https://szagoruyko.github.io/ Sergey Zagoruyko]

Семинарист: [https://www.hse.ru/org/persons/401628708/ Eva Neudachina]

Локация: лекция - G002, семинар - R208.

Материалы: [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/lecture_06.mp4 запись лекции], [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/seminar_06.mp4 запись семинара], [https://github.com/thecrazymage/DL2_HSE/tree/main/week_06 материалы].

'''Лекция / Семинар 7 (20.10).'''

Тема: Diffusion models 1

Аннотация лекции и семинара: Tomorrow's lecture and seminar will be devoted to an introduction to diffusion models. Diffusion models are currently the most popular approach to generative modeling due to their high-quality generation and diversity (mode coverage) of the learned distribution. The idea behind diffusion models is to consider the process of gradually transforming data into pure noise and construct its inverse in time, which will transform noise into data. In the lecture and seminar, we will work with noise processes and derive the classic [https://arxiv.org/abs/2006.11239 DDPM] model, which proposes to minimize the KL-divergence between the “true” reverse process that converts noise into data and the denoising process specified by the neural network. In the process, we will see that this procedure is equivalent to training a denoiser neural network that predicts a clean object from a noisy one. In addition, we will interpret the resulting denoising process: in it, each step corresponds to replacing part of the current noisy image with an (increasingly high-quality) prediction of the denoiser.

Лектор и семинарист: [https://www.hse.ru/org/persons/190910999/ Denis Rakitin]

Локация: лекция - G002, семинар - R208.

Материалы: [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/lecture_07.mp4 запись лекции], [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/seminar_07.mp4 запись семинара], [https://github.com/thecrazymage/DL2_HSE/tree/main/week_07 материалы].

'''Лекция / Семинар 8 (03.11).'''

Тема: Diffusion models 2

Аннотация лекции и семинара: In this lecture and seminar, we will continue our exploration of diffusion models. We will introduce the score function and score identity, present classifier and classifier-free guidance, and derive DDIM model.

Лектор и семинарист: [https://www.hse.ru/org/persons/190910999/ Denis Rakitin]

Локация: лекция - R205, семинар - R208.

Материалы: [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/lecture_08.mp4 запись лекции], [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/seminar_08.mp4 запись семинара], [https://github.com/thecrazymage/DL2_HSE/tree/main/week_08 материалы].

'''Лекция / Семинар 9 (10.11).'''

Тема: Diffusion models 3

Аннотация лекции: This lecture presents applied aspects of diffusion models, a class of generative methods that have demonstrated strong performance across images, text, and audio. We will review modern architectures, with a focus on [https://github.com/black-forest-labs/flux FLUX], and unpack the key principles of their design and training.

Лектор: [https://scholar.google.com/citations?user=o6pRm_gAAAAJ&hl=ru Nikita Starodubcev]

Аннотация семинара: The seminar will provide a hands-on, in-depth inspection of [https://github.com/PixArt-alpha/PixArt-alpha PixArt] model. Then ee will attempt further fine-tuning via [https://dreambooth.github.io/ DreamBooth], review common evaluation metrics, and compare different models and configurations.

Семинарист: [https://www.hse.ru/org/persons/401628708/ Eva Neudachina]

Локация: лекция - R205, семинар - R208.

Материалы: [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/lecture_09.mp4 запись лекции], [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/seminar_09.mp4 запись семинара], [https://github.com/thecrazymage/DL2_HSE/tree/main/week_09 материалы].

'''Лекция / Семинар 10 (17.11).'''

Тема: 3D CV

Аннотация лекции и семинара: This lecture and seminar will discuss how diffusion models can be used for 3D computer vision tasks such as generation and reconstruction, and how they relate to methods like NeRF and Gaussian splatting. We will also look at how these models can be applied to problems like novel view synthesis, relighting, and camera pose or depth estimation.

Лектор и семинарист: [https://www.hse.ru/org/persons/885876805/ Mishan Aliev]

Локация: лекция - R205, семинар - R208.

Материалы: [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/lecture_10.mp4 запись лекции], [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/seminar_10.mp4 запись семинара], [https://github.com/thecrazymage/DL2_HSE/tree/main/week_10 материалы].

'''Лекция / Семинар 11 (24.11).'''

Тема: Neural Recommender Systems

Аннотация лекции и семинара: Modern recommender systems also make heavy use of deep learning — both discriminative and generative models. A system has to understand content (tracks, products, videos, etc.), model user behavior (short- and long-term preferences), and predict which content each user is likely to enjoy. At the same time, recommender systems come with their own challenges: cold start for new items and users, huge billion-scale item catalogs with heavy-tailed popularity distributions (popularity bias), and constant distribution drift, where all the underlying distributions keep changing along with the rest of the world. We will discuss: 1) Two-tower neural networks for candidate generation; 2) Sequential recommendation and more modern approaches to generative modeling of users; 3) Semantic IDs and tuning LLMs for recommendation 4) What makes the recommendation domain different from other deep learning domains.

Лектор: [https://www.hse.ru/org/persons/1085421573/ Kirill Khrylchenko]

Семинарист: [https://www.linkedin.com/in/artem-matveev-7b2725255/ Artem Matveev]

Локация: лекция - R205, семинар - R208.

Материалы: [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/lecture_11.mp4 запись лекции], [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/seminar_11.mp4 запись семинара], [https://github.com/thecrazymage/DL2_HSE/tree/main/week_11 материалы].

'''Лекция / Семинар 12 (08.12).'''

Тема: Graph Machine Learning

Аннотация лекции и семинара: In the lecture, we will explore the basics of Graph Machine Learning (Graph-ML), the tasks are being solved in this domain, and the properties of the models and architectures being used in Graph-ML. In the seminar, we will train our Graph Neural Network using modern Graph-ML frameworks.

Лектор и семинарист: [https://www.hse.ru/org/persons/816100677/ Fedor Velikonivtsev]

Локация: лекция - R205, семинар - R208.

Материалы: tba.

== Домашние задания ==

{| class="wikitable"
! № !! Домашнее задание !! Ссылка !! Дедлайн (жёсткий)
|-
| 1 || Tensor and DL Libraries || https://github.com/thecrazymage/DL2_HSE/tree/main/homeworks/homework_01 || 30 сентября, 23:59
|-
| 2 || Transformers for Named Entity Recognition || https://github.com/thecrazymage/DL2_HSE/tree/main/homeworks/homework_02 || 14 октября, 23:59
|-
| 3 || Image Segmentation || https://github.com/thecrazymage/DL2_HSE/tree/main/homeworks/homework_03 || 2 ноября, 23:59
|-
| 4 || Diffusion Models || https://github.com/thecrazymage/DL2_HSE/tree/main/homeworks/homework_04 || 18 ноября, 23:59
|-
| 5 || Retrieval || https://github.com/thecrazymage/DL2_HSE/tree/main/homeworks/homework_05 || 17 декабря, 23:59

|}

== Соревнование ==

'''Задание:''' https://github.com/thecrazymage/DL2_HSE/tree/main/homeworks/competition

'''Возможность перезачета:''' https://t.me/c/2969026465/2/1601.

'''Дедлайн (жёсткий):''' 19 декабря 2025, 23:59

Глубинное обучение 2 2025

2025-12-08T13:51:13Z

Anapluslap: add class 12 announcement

== Общая информация ==
Курс предназначен для студентов 4 курса ФКН ПМИ (МОП и КНАД).

Занятия проходят '''по понедельникам 14:40-17:40''' (переносы будут сообщаться в чате).

Полезные ссылки:
* Чат с обсуждением: https://t.me/+8dcjl4gHlyEwYzcy
* Репозиторий курса: https://github.com/thecrazymage/DL2_HSE
* Таблица с оценками: https://docs.google.com/spreadsheets/d/18rfZbf7Zsm-xmbiQn_eJVDXXbDfesbHT/edit?usp=sharing&ouid=115132401804687564737&rtpof=true&sd=true
* Anytask: https://anytask.org/course/1219

Формула итоговой оценки (округление арифметическое):
# МОП: Оитог = 0.25 * Осоревнование + 0.75 * ОДЗ,
# КНАД: Оитог = ОДЗ,
где ОДЗ - средняя оценка за практические домашние задания.

== Преподаватели и ассистенты ==

Кому писать, если кажется, что все пропало: [https://t.me/MishanAliev Мишан Алиев]

{| class="wikitable"
|-
! Группа !! Семинарист !! Ассистенты !! Чаты групп !! Инвайт в anytask
|-
| 221 (МОП) || [https://t.me/mightyneighbor Федя Великонивцев] || [https://t.me/vsem_paket Динар Саберов], [https://t.me/annvasileeva Анна Василева] || [https://t.me/+XzlFh0Y8GHs3NTMy МОП 221] || yvZZTIS
|-
| 222 (МОП) || [https://t.me/cocosinca Ева Неудачина] || [https://t.me/mat_os Александр Матосян], [https://t.me/kade1shvili Полина Кадейшвили] || [https://t.me/+PpxiIpF7wLYyMjE6 МОП 222] || W38CfZf
|-
| 223 (МОП) || [https://t.me/tutugarin Иван Ершов] || [https://t.me/Andrew_ut Андрей Уткин], [https://t.me/uzhedevyat Георгий Фатахов] || [https://t.me/+xRZkDO4TjiA1YjQ6 МОП 223] || ph91Jlz
|-
| 224 (МОП) || [https://t.me/Stepyndriy Степан Беляков] || [https://t.me/ponomarchuk_anna Анна Пономарчук], [https://t.me/bogsolntca Татьяна Яковлева] || [https://t.me/+XtcOG352A-IxMmY6 МОП 224] || iPwd342
|-
| КНАД || [https://t.me/burakevi4 Даня Бураков] || [https://t.me/anapluslap Анастасия Лапшина], [https://t.me/irbix7 Иван Галий] || [https://t.me/+JNPUgVxh6ophOGQy КНАД] || zTn4sRP
|}

== Лекции и семинары ==

'''Лекция / Семинар 1 (08.09).'''

Тема: Essentials of GPU, Deep Learning Bottlenecks, and Benchmarking Basics

Аннотация: In this session, we will explore the reasons behind the dominance of GPUs in Deep Learning and examine the common sources of performance bottlenecks in DL code. You will learn how to identify these bottlenecks using profiling tools and apply techniques to optimize and accelerate your code.

Лектор и семинарист: [https://www.hse.ru/org/persons/816100677/ Fedor Velikonivtsev]

Локация: лекция - R208, семинар - D102.

Материалы: [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/lecture_01.mp4 запись лекции], [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/seminar_01.mp4 запись семинара], [https://github.com/thecrazymage/DL2_HSE/tree/main/week_01 материалы].

'''Лекция / Семинар 2 (15.09).'''

Тема: On Transformers and Bitter Lesson

Аннотация: In this talk we’ll dive into the landscape of model architectures in deep learning, with a focus on the world around transformers. We’ll briefly recall what a transformer is, trace the evolution from encoder–decoder to encoder-only and decoder-only models, and touch on the rise of “efficient mixers” such as state space models, linear attention, and beyond. We’ll conclude by reflecting on the role of data and compute. The lecture is inspired in part by [http://incompleteideas.net/IncIdeas/BitterLesson.html The Bitter Lesson] and blends a brain dump with some entertaining insights from recent years of architectural exploration.

Лектор и семинарист: [https://www.hse.ru/org/persons/190912012/ Ivan Rubachev]

Локация: лекция - R206, семинар - R208.

Материалы: [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/lecture%2Bseminar_02.mp4 запись лекции и семинара], [https://github.com/thecrazymage/DL2_HSE/tree/main/week_02 материалы].

'''Лекция / Семинар 3 (22.09).'''

Тема: Modern LLMs essentials

Аннотация лекции: This week, we will discuss LLMs. We will discuss why they are so effective for text generation, how they can be applied to different NLP problems, and the risks they pose. You will learn the details of RLHF, PEFT, and RAG, which make LLMs robust in various cases.

Лектор: [https://www.hse.ru/org/persons/208533329/ Alexander Shabalin]

Локация: онлайн.

Аннотация семинара: In this seminar, we will explore the concept of LLM-based agents and how they extend the capabilities of modern language models. We will discuss function calling as a way to integrate external tools, chain-of-thought reasoning for structured problem solving, and reinforcement learning techniques for training agents.

Семинарист: [http://www.instagram.com/tugarin_vanya Ivan Ershov]

Локация: R208.

Материалы: [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/lecture_03.mp4 запись лекции], [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/seminar_03.mp4 запись семинара], [https://github.com/thecrazymage/DL2_HSE/tree/main/week_03 материалы].

'''Лекция / Семинар 4 (29.09).'''

Тема: Basics of Efficient LLM Training Infrastructure

Аннотация: In this lecture, we will study the basic rules that underpin the infrastructure for efficient large language model training. We will also examine common problems that may arise in this process and explore practical ways to address them.

Лектор и семинарист: [https://habr.com/ru/users/MichaelEk/ Michael Khrushchev]

Локация: лекция - G002, семинар - R208.

Материалы: [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/lecture%2Bseminar_04.mp4 запись лекции и семинара], [https://github.com/thecrazymage/DL2_HSE/tree/main/week_04 материалы].

'''Лекция / Семинар 5 (06.10).'''

Тема: Segmentation and Detection

Аннотация лекции: This week, we'll explore the evolution of object detection and segmentation — from R-CNN to real-time methods like YOLO-World and CLIP-based approaches. We'll examine how U-Net architectures have transcended computer vision to power neural operators and how to use diffusion models for segmentation tasks.

Аннотация семинара: We will implement architectures and train a semantic segmentation model. We will discuss regularization methods for convolutional layers. In addition, we will learn how to use pre-trained models and apply them.

Лектор и семинарист: [https://www.hse.ru/org/persons/956617478/ Alexander Oganov]

Локация: лекция - G002, семинар - R208.

Материалы: [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/lecture_05.mp4 запись лекции], [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/seminar_05.mp4 запись семинара], [https://github.com/thecrazymage/DL2_HSE/tree/main/week_05 материалы].

'''Лекция / Семинар 6 (13.10).'''

Тема: Segmentation and Detection 2

Аннотация лекции и семинара: In this lecture, we will continue in more detail about segmentation and detection.

Лектор: [https://szagoruyko.github.io/ Sergey Zagoruyko]

Семинарист: [https://www.hse.ru/org/persons/401628708/ Eva Neudachina]

Локация: лекция - G002, семинар - R208.

Материалы: [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/lecture_06.mp4 запись лекции], [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/seminar_06.mp4 запись семинара], [https://github.com/thecrazymage/DL2_HSE/tree/main/week_06 материалы].

'''Лекция / Семинар 7 (20.10).'''

Тема: Diffusion models 1

Аннотация лекции и семинара: Tomorrow's lecture and seminar will be devoted to an introduction to diffusion models. Diffusion models are currently the most popular approach to generative modeling due to their high-quality generation and diversity (mode coverage) of the learned distribution. The idea behind diffusion models is to consider the process of gradually transforming data into pure noise and construct its inverse in time, which will transform noise into data. In the lecture and seminar, we will work with noise processes and derive the classic [https://arxiv.org/abs/2006.11239 DDPM] model, which proposes to minimize the KL-divergence between the “true” reverse process that converts noise into data and the denoising process specified by the neural network. In the process, we will see that this procedure is equivalent to training a denoiser neural network that predicts a clean object from a noisy one. In addition, we will interpret the resulting denoising process: in it, each step corresponds to replacing part of the current noisy image with an (increasingly high-quality) prediction of the denoiser.

Лектор и семинарист: [https://www.hse.ru/org/persons/190910999/ Denis Rakitin]

Локация: лекция - G002, семинар - R208.

Материалы: [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/lecture_07.mp4 запись лекции], [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/seminar_07.mp4 запись семинара], [https://github.com/thecrazymage/DL2_HSE/tree/main/week_07 материалы].

'''Лекция / Семинар 8 (03.11).'''

Тема: Diffusion models 2

Аннотация лекции и семинара: In this lecture and seminar, we will continue our exploration of diffusion models. We will introduce the score function and score identity, present classifier and classifier-free guidance, and derive DDIM model.

Лектор и семинарист: [https://www.hse.ru/org/persons/190910999/ Denis Rakitin]

Локация: лекция - R205, семинар - R208.

Материалы: [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/lecture_08.mp4 запись лекции], [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/seminar_08.mp4 запись семинара], [https://github.com/thecrazymage/DL2_HSE/tree/main/week_08 материалы].

'''Лекция / Семинар 9 (10.11).'''

Тема: Diffusion models 3

Аннотация лекции: This lecture presents applied aspects of diffusion models, a class of generative methods that have demonstrated strong performance across images, text, and audio. We will review modern architectures, with a focus on [https://github.com/black-forest-labs/flux FLUX], and unpack the key principles of their design and training.

Лектор: [https://scholar.google.com/citations?user=o6pRm_gAAAAJ&hl=ru Nikita Starodubcev]

Аннотация семинара: The seminar will provide a hands-on, in-depth inspection of [https://github.com/PixArt-alpha/PixArt-alpha PixArt] model. Then ee will attempt further fine-tuning via [https://dreambooth.github.io/ DreamBooth], review common evaluation metrics, and compare different models and configurations.

Семинарист: [https://www.hse.ru/org/persons/401628708/ Eva Neudachina]

Локация: лекция - R205, семинар - R208.

Материалы: [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/lecture_09.mp4 запись лекции], [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/seminar_09.mp4 запись семинара], [https://github.com/thecrazymage/DL2_HSE/tree/main/week_09 материалы].

'''Лекция / Семинар 10 (17.11).'''

Тема: 3D CV

Аннотация лекции и семинара: This lecture and seminar will discuss how diffusion models can be used for 3D computer vision tasks such as generation and reconstruction, and how they relate to methods like NeRF and Gaussian splatting. We will also look at how these models can be applied to problems like novel view synthesis, relighting, and camera pose or depth estimation.

Лектор и семинарист: [https://www.hse.ru/org/persons/885876805/ Mishan Aliev]

Локация: лекция - R205, семинар - R208.

Материалы: [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/lecture_10.mp4 запись лекции], [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/seminar_10.mp4 запись семинара], [https://github.com/thecrazymage/DL2_HSE/tree/main/week_10 материалы].

'''Лекция / Семинар 11 (24.11).'''

Тема: Neural Recommender Systems

Аннотация лекции и семинара: Modern recommender systems also make heavy use of deep learning — both discriminative and generative models. A system has to understand content (tracks, products, videos, etc.), model user behavior (short- and long-term preferences), and predict which content each user is likely to enjoy. At the same time, recommender systems come with their own challenges: cold start for new items and users, huge billion-scale item catalogs with heavy-tailed popularity distributions (popularity bias), and constant distribution drift, where all the underlying distributions keep changing along with the rest of the world. We will discuss: 1) Two-tower neural networks for candidate generation; 2) Sequential recommendation and more modern approaches to generative modeling of users; 3) Semantic IDs and tuning LLMs for recommendation 4) What makes the recommendation domain different from other deep learning domains.

Лектор: [https://www.hse.ru/org/persons/1085421573/ Kirill Khrylchenko]

Семинарист: [https://www.linkedin.com/in/artem-matveev-7b2725255/ Artem Matveev]

Локация: лекция - R205, семинар - R208.

Материалы: [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/lecture_11.mp4 запись лекции], [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/seminar_11.mp4 запись семинара], [https://github.com/thecrazymage/DL2_HSE/tree/main/week_11 материалы].

'''Лекция / Семинар 12 (08.12).'''

Тема: Graph Machine Learning

Аннотация лекции и семинара: In the lecture, we will explore the basics of Graph Machine Learning (Graph-ML), the tasks are being solved in this domain, and the properties of the models and architectures being used in Graph-ML. In the seminar, we will train our Graph Neural Network using modern Graph-ML frameworks.

Лектор и семинарист: [https://www.hse.ru/org/persons/816100677/ Fedor Velikonivtsev]

Локация: лекция - R205, семинар - R208.

Материалы: tba.

== Домашние задания ==

{| class="wikitable"
! № !! Домашнее задание !! Ссылка !! Дедлайн (жёсткий)
|-
| 1 || Tensor and DL Libraries || https://github.com/thecrazymage/DL2_HSE/tree/main/homeworks/homework_01 || 30 сентября, 23:59
|-
| 2 || Transformers for Named Entity Recognition || https://github.com/thecrazymage/DL2_HSE/tree/main/homeworks/homework_02 || 14 октября, 23:59
|-
| 3 || Image Segmentation || https://github.com/thecrazymage/DL2_HSE/tree/main/homeworks/homework_03 || 2 ноября, 23:59
|-
| 4 || Diffusion Models || https://github.com/thecrazymage/DL2_HSE/tree/main/homeworks/homework_04 || 18 ноября, 23:59
|-
| 5 || Retrieval || https://github.com/thecrazymage/DL2_HSE/tree/main/homeworks/homework_05 || 10 декабря, 23:59

|}

== Соревнование ==

'''Задание:''' https://github.com/thecrazymage/DL2_HSE/tree/main/homeworks/competition

'''Возможность перезачета:''' https://t.me/c/2969026465/2/1601.

'''Дедлайн (жёсткий):''' 19 декабря 2025, 23:59

Глубинное обучение 2 2025

2025-12-02T16:14:08Z

Anapluslap: add competition

== Общая информация ==
Курс предназначен для студентов 4 курса ФКН ПМИ (МОП и КНАД).

Занятия проходят '''по понедельникам 14:40-17:40''' (переносы будут сообщаться в чате).

Полезные ссылки:
* Чат с обсуждением: https://t.me/+8dcjl4gHlyEwYzcy
* Репозиторий курса: https://github.com/thecrazymage/DL2_HSE
* Таблица с оценками: https://docs.google.com/spreadsheets/d/18rfZbf7Zsm-xmbiQn_eJVDXXbDfesbHT/edit?usp=sharing&ouid=115132401804687564737&rtpof=true&sd=true
* Anytask: https://anytask.org/course/1219

Формула итоговой оценки (округление арифметическое):
# МОП: Оитог = 0.25 * Осоревнование + 0.75 * ОДЗ,
# КНАД: Оитог = ОДЗ,
где ОДЗ - средняя оценка за практические домашние задания.

== Преподаватели и ассистенты ==

Кому писать, если кажется, что все пропало: [https://t.me/MishanAliev Мишан Алиев]

{| class="wikitable"
|-
! Группа !! Семинарист !! Ассистенты !! Чаты групп !! Инвайт в anytask
|-
| 221 (МОП) || [https://t.me/mightyneighbor Федя Великонивцев] || [https://t.me/vsem_paket Динар Саберов], [https://t.me/annvasileeva Анна Василева] || [https://t.me/+XzlFh0Y8GHs3NTMy МОП 221] || yvZZTIS
|-
| 222 (МОП) || [https://t.me/cocosinca Ева Неудачина] || [https://t.me/mat_os Александр Матосян], [https://t.me/kade1shvili Полина Кадейшвили] || [https://t.me/+PpxiIpF7wLYyMjE6 МОП 222] || W38CfZf
|-
| 223 (МОП) || [https://t.me/tutugarin Иван Ершов] || [https://t.me/Andrew_ut Андрей Уткин], [https://t.me/uzhedevyat Георгий Фатахов] || [https://t.me/+xRZkDO4TjiA1YjQ6 МОП 223] || ph91Jlz
|-
| 224 (МОП) || [https://t.me/Stepyndriy Степан Беляков] || [https://t.me/ponomarchuk_anna Анна Пономарчук], [https://t.me/bogsolntca Татьяна Яковлева] || [https://t.me/+XtcOG352A-IxMmY6 МОП 224] || iPwd342
|-
| КНАД || [https://t.me/burakevi4 Даня Бураков] || [https://t.me/anapluslap Анастасия Лапшина], [https://t.me/irbix7 Иван Галий] || [https://t.me/+JNPUgVxh6ophOGQy КНАД] || zTn4sRP
|}

== Лекции и семинары ==

'''Лекция / Семинар 1 (08.09).'''

Тема: Essentials of GPU, Deep Learning Bottlenecks, and Benchmarking Basics

Аннотация: In this session, we will explore the reasons behind the dominance of GPUs in Deep Learning and examine the common sources of performance bottlenecks in DL code. You will learn how to identify these bottlenecks using profiling tools and apply techniques to optimize and accelerate your code.

Лектор и семинарист: [https://www.hse.ru/org/persons/816100677/ Fedor Velikonivtsev]

Локация: лекция - R208, семинар - D102.

Материалы: [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/lecture_01.mp4 запись лекции], [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/seminar_01.mp4 запись семинара], [https://github.com/thecrazymage/DL2_HSE/tree/main/week_01 материалы].

'''Лекция / Семинар 2 (15.09).'''

Тема: On Transformers and Bitter Lesson

Аннотация: In this talk we’ll dive into the landscape of model architectures in deep learning, with a focus on the world around transformers. We’ll briefly recall what a transformer is, trace the evolution from encoder–decoder to encoder-only and decoder-only models, and touch on the rise of “efficient mixers” such as state space models, linear attention, and beyond. We’ll conclude by reflecting on the role of data and compute. The lecture is inspired in part by [http://incompleteideas.net/IncIdeas/BitterLesson.html The Bitter Lesson] and blends a brain dump with some entertaining insights from recent years of architectural exploration.

Лектор и семинарист: [https://www.hse.ru/org/persons/190912012/ Ivan Rubachev]

Локация: лекция - R206, семинар - R208.

Материалы: [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/lecture%2Bseminar_02.mp4 запись лекции и семинара], [https://github.com/thecrazymage/DL2_HSE/tree/main/week_02 материалы].

'''Лекция / Семинар 3 (22.09).'''

Тема: Modern LLMs essentials

Аннотация лекции: This week, we will discuss LLMs. We will discuss why they are so effective for text generation, how they can be applied to different NLP problems, and the risks they pose. You will learn the details of RLHF, PEFT, and RAG, which make LLMs robust in various cases.

Лектор: [https://www.hse.ru/org/persons/208533329/ Alexander Shabalin]

Локация: онлайн.

Аннотация семинара: In this seminar, we will explore the concept of LLM-based agents and how they extend the capabilities of modern language models. We will discuss function calling as a way to integrate external tools, chain-of-thought reasoning for structured problem solving, and reinforcement learning techniques for training agents.

Семинарист: [http://www.instagram.com/tugarin_vanya Ivan Ershov]

Локация: R208.

Материалы: [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/lecture_03.mp4 запись лекции], [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/seminar_03.mp4 запись семинара], [https://github.com/thecrazymage/DL2_HSE/tree/main/week_03 материалы].

'''Лекция / Семинар 4 (29.09).'''

Тема: Basics of Efficient LLM Training Infrastructure

Аннотация: In this lecture, we will study the basic rules that underpin the infrastructure for efficient large language model training. We will also examine common problems that may arise in this process and explore practical ways to address them.

Лектор и семинарист: [https://habr.com/ru/users/MichaelEk/ Michael Khrushchev]

Локация: лекция - G002, семинар - R208.

Материалы: [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/lecture%2Bseminar_04.mp4 запись лекции и семинара], [https://github.com/thecrazymage/DL2_HSE/tree/main/week_04 материалы].

'''Лекция / Семинар 5 (06.10).'''

Тема: Segmentation and Detection

Аннотация лекции: This week, we'll explore the evolution of object detection and segmentation — from R-CNN to real-time methods like YOLO-World and CLIP-based approaches. We'll examine how U-Net architectures have transcended computer vision to power neural operators and how to use diffusion models for segmentation tasks.

Аннотация семинара: We will implement architectures and train a semantic segmentation model. We will discuss regularization methods for convolutional layers. In addition, we will learn how to use pre-trained models and apply them.

Лектор и семинарист: [https://www.hse.ru/org/persons/956617478/ Alexander Oganov]

Локация: лекция - G002, семинар - R208.

Материалы: [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/lecture_05.mp4 запись лекции], [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/seminar_05.mp4 запись семинара], [https://github.com/thecrazymage/DL2_HSE/tree/main/week_05 материалы].

'''Лекция / Семинар 6 (13.10).'''

Тема: Segmentation and Detection 2

Аннотация лекции и семинара: In this lecture, we will continue in more detail about segmentation and detection.

Лектор: [https://szagoruyko.github.io/ Sergey Zagoruyko]

Семинарист: [https://www.hse.ru/org/persons/401628708/ Eva Neudachina]

Локация: лекция - G002, семинар - R208.

Материалы: [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/lecture_06.mp4 запись лекции], [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/seminar_06.mp4 запись семинара], [https://github.com/thecrazymage/DL2_HSE/tree/main/week_06 материалы].

'''Лекция / Семинар 7 (20.10).'''

Тема: Diffusion models 1

Аннотация лекции и семинара: Tomorrow's lecture and seminar will be devoted to an introduction to diffusion models. Diffusion models are currently the most popular approach to generative modeling due to their high-quality generation and diversity (mode coverage) of the learned distribution. The idea behind diffusion models is to consider the process of gradually transforming data into pure noise and construct its inverse in time, which will transform noise into data. In the lecture and seminar, we will work with noise processes and derive the classic [https://arxiv.org/abs/2006.11239 DDPM] model, which proposes to minimize the KL-divergence between the “true” reverse process that converts noise into data and the denoising process specified by the neural network. In the process, we will see that this procedure is equivalent to training a denoiser neural network that predicts a clean object from a noisy one. In addition, we will interpret the resulting denoising process: in it, each step corresponds to replacing part of the current noisy image with an (increasingly high-quality) prediction of the denoiser.

Лектор и семинарист: [https://www.hse.ru/org/persons/190910999/ Denis Rakitin]

Локация: лекция - G002, семинар - R208.

Материалы: [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/lecture_07.mp4 запись лекции], [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/seminar_07.mp4 запись семинара], [https://github.com/thecrazymage/DL2_HSE/tree/main/week_07 материалы].

'''Лекция / Семинар 8 (03.11).'''

Тема: Diffusion models 2

Аннотация лекции и семинара: In this lecture and seminar, we will continue our exploration of diffusion models. We will introduce the score function and score identity, present classifier and classifier-free guidance, and derive DDIM model.

Лектор и семинарист: [https://www.hse.ru/org/persons/190910999/ Denis Rakitin]

Локация: лекция - R205, семинар - R208.

Материалы: [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/lecture_08.mp4 запись лекции], [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/seminar_08.mp4 запись семинара], [https://github.com/thecrazymage/DL2_HSE/tree/main/week_08 материалы].

'''Лекция / Семинар 9 (10.11).'''

Тема: Diffusion models 3

Аннотация лекции: This lecture presents applied aspects of diffusion models, a class of generative methods that have demonstrated strong performance across images, text, and audio. We will review modern architectures, with a focus on [https://github.com/black-forest-labs/flux FLUX], and unpack the key principles of their design and training.

Лектор: [https://scholar.google.com/citations?user=o6pRm_gAAAAJ&hl=ru Nikita Starodubcev]

Аннотация семинара: The seminar will provide a hands-on, in-depth inspection of [https://github.com/PixArt-alpha/PixArt-alpha PixArt] model. Then ee will attempt further fine-tuning via [https://dreambooth.github.io/ DreamBooth], review common evaluation metrics, and compare different models and configurations.

Семинарист: [https://www.hse.ru/org/persons/401628708/ Eva Neudachina]

Локация: лекция - R205, семинар - R208.

Материалы: [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/lecture_09.mp4 запись лекции], [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/seminar_09.mp4 запись семинара], [https://github.com/thecrazymage/DL2_HSE/tree/main/week_09 материалы].

'''Лекция / Семинар 10 (17.11).'''

Тема: 3D CV

Аннотация лекции и семинара: This lecture and seminar will discuss how diffusion models can be used for 3D computer vision tasks such as generation and reconstruction, and how they relate to methods like NeRF and Gaussian splatting. We will also look at how these models can be applied to problems like novel view synthesis, relighting, and camera pose or depth estimation.

Лектор и семинарист: [https://www.hse.ru/org/persons/885876805/ Mishan Aliev]

Локация: лекция - R205, семинар - R208.

Материалы: [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/lecture_10.mp4 запись лекции], [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/seminar_10.mp4 запись семинара], [https://github.com/thecrazymage/DL2_HSE/tree/main/week_10 материалы].

'''Лекция / Семинар 11 (24.11).'''

Тема: Neural Recommender Systems

Аннотация лекции и семинара: Modern recommender systems also make heavy use of deep learning — both discriminative and generative models. A system has to understand content (tracks, products, videos, etc.), model user behavior (short- and long-term preferences), and predict which content each user is likely to enjoy. At the same time, recommender systems come with their own challenges: cold start for new items and users, huge billion-scale item catalogs with heavy-tailed popularity distributions (popularity bias), and constant distribution drift, where all the underlying distributions keep changing along with the rest of the world. We will discuss: 1) Two-tower neural networks for candidate generation; 2) Sequential recommendation and more modern approaches to generative modeling of users; 3) Semantic IDs and tuning LLMs for recommendation 4) What makes the recommendation domain different from other deep learning domains.

Лектор: [https://www.hse.ru/org/persons/1085421573/ Kirill Khrylchenko]

Семинарист: [https://www.linkedin.com/in/artem-matveev-7b2725255/ Artem Matveev]

Локация: лекция - R205, семинар - R208.

Материалы: [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/lecture_11.mp4 запись лекции], [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/seminar_11.mp4 запись семинара], [https://github.com/thecrazymage/DL2_HSE/tree/main/week_11 материалы].

== Домашние задания ==

{| class="wikitable"
! № !! Домашнее задание !! Ссылка !! Дедлайн (жёсткий)
|-
| 1 || Tensor and DL Libraries || https://github.com/thecrazymage/DL2_HSE/tree/main/homeworks/homework_01 || 30 сентября, 23:59
|-
| 2 || Transformers for Named Entity Recognition || https://github.com/thecrazymage/DL2_HSE/tree/main/homeworks/homework_02 || 14 октября, 23:59
|-
| 3 || Image Segmentation || https://github.com/thecrazymage/DL2_HSE/tree/main/homeworks/homework_03 || 2 ноября, 23:59
|-
| 4 || Diffusion Models || https://github.com/thecrazymage/DL2_HSE/tree/main/homeworks/homework_04 || 18 ноября, 23:59
|-
| 5 || Retrieval || https://github.com/thecrazymage/DL2_HSE/tree/main/homeworks/homework_05 || 10 декабря, 23:59

|}

== Соревнование ==

'''Задание:''' https://github.com/thecrazymage/DL2_HSE/tree/main/homeworks/competition

'''Возможность перезачета:''' https://t.me/c/2969026465/2/1601.

'''Дедлайн (жёсткий):''' 19 декабря 2025, 23:59

Глубинное обучение 2 2025

2025-11-26T17:22:36Z

Anapluslap: add hw 5

== Общая информация ==
Курс предназначен для студентов 4 курса ФКН ПМИ (МОП и КНАД).

Занятия проходят '''по понедельникам 14:40-17:40''' (переносы будут сообщаться в чате).

Полезные ссылки:
* Чат с обсуждением: https://t.me/+8dcjl4gHlyEwYzcy
* Репозиторий курса: https://github.com/thecrazymage/DL2_HSE
* Таблица с оценками: https://docs.google.com/spreadsheets/d/18rfZbf7Zsm-xmbiQn_eJVDXXbDfesbHT/edit?usp=sharing&ouid=115132401804687564737&rtpof=true&sd=true
* Anytask: https://anytask.org/course/1219

Формула итоговой оценки (округление арифметическое):
# МОП: Оитог = 0.25 * Осоревнование + 0.75 * ОДЗ,
# КНАД: Оитог = ОДЗ,
где ОДЗ - средняя оценка за практические домашние задания.

== Преподаватели и ассистенты ==

Кому писать, если кажется, что все пропало: [https://t.me/MishanAliev Мишан Алиев]

{| class="wikitable"
|-
! Группа !! Семинарист !! Ассистенты !! Чаты групп !! Инвайт в anytask
|-
| 221 (МОП) || [https://t.me/mightyneighbor Федя Великонивцев] || [https://t.me/vsem_paket Динар Саберов], [https://t.me/annvasileeva Анна Василева] || [https://t.me/+XzlFh0Y8GHs3NTMy МОП 221] || yvZZTIS
|-
| 222 (МОП) || [https://t.me/cocosinca Ева Неудачина] || [https://t.me/mat_os Александр Матосян], [https://t.me/kade1shvili Полина Кадейшвили] || [https://t.me/+PpxiIpF7wLYyMjE6 МОП 222] || W38CfZf
|-
| 223 (МОП) || [https://t.me/tutugarin Иван Ершов] || [https://t.me/Andrew_ut Андрей Уткин], [https://t.me/uzhedevyat Георгий Фатахов] || [https://t.me/+xRZkDO4TjiA1YjQ6 МОП 223] || ph91Jlz
|-
| 224 (МОП) || [https://t.me/Stepyndriy Степан Беляков] || [https://t.me/ponomarchuk_anna Анна Пономарчук], [https://t.me/bogsolntca Татьяна Яковлева] || [https://t.me/+XtcOG352A-IxMmY6 МОП 224] || iPwd342
|-
| КНАД || [https://t.me/burakevi4 Даня Бураков] || [https://t.me/anapluslap Анастасия Лапшина], [https://t.me/irbix7 Иван Галий] || [https://t.me/+JNPUgVxh6ophOGQy КНАД] || zTn4sRP
|}

== Лекции и семинары ==

'''Лекция / Семинар 1 (08.09).'''

Тема: Essentials of GPU, Deep Learning Bottlenecks, and Benchmarking Basics

Аннотация: In this session, we will explore the reasons behind the dominance of GPUs in Deep Learning and examine the common sources of performance bottlenecks in DL code. You will learn how to identify these bottlenecks using profiling tools and apply techniques to optimize and accelerate your code.

Лектор и семинарист: [https://www.hse.ru/org/persons/816100677/ Fedor Velikonivtsev]

Локация: лекция - R208, семинар - D102.

Материалы: [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/lecture_01.mp4 запись лекции], [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/seminar_01.mp4 запись семинара], [https://github.com/thecrazymage/DL2_HSE/tree/main/week_01 материалы].

'''Лекция / Семинар 2 (15.09).'''

Тема: On Transformers and Bitter Lesson

Аннотация: In this talk we’ll dive into the landscape of model architectures in deep learning, with a focus on the world around transformers. We’ll briefly recall what a transformer is, trace the evolution from encoder–decoder to encoder-only and decoder-only models, and touch on the rise of “efficient mixers” such as state space models, linear attention, and beyond. We’ll conclude by reflecting on the role of data and compute. The lecture is inspired in part by [http://incompleteideas.net/IncIdeas/BitterLesson.html The Bitter Lesson] and blends a brain dump with some entertaining insights from recent years of architectural exploration.

Лектор и семинарист: [https://www.hse.ru/org/persons/190912012/ Ivan Rubachev]

Локация: лекция - R206, семинар - R208.

Материалы: [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/lecture%2Bseminar_02.mp4 запись лекции и семинара], [https://github.com/thecrazymage/DL2_HSE/tree/main/week_02 материалы].

'''Лекция / Семинар 3 (22.09).'''

Тема: Modern LLMs essentials

Аннотация лекции: This week, we will discuss LLMs. We will discuss why they are so effective for text generation, how they can be applied to different NLP problems, and the risks they pose. You will learn the details of RLHF, PEFT, and RAG, which make LLMs robust in various cases.

Лектор: [https://www.hse.ru/org/persons/208533329/ Alexander Shabalin]

Локация: онлайн.

Аннотация семинара: In this seminar, we will explore the concept of LLM-based agents and how they extend the capabilities of modern language models. We will discuss function calling as a way to integrate external tools, chain-of-thought reasoning for structured problem solving, and reinforcement learning techniques for training agents.

Семинарист: [http://www.instagram.com/tugarin_vanya Ivan Ershov]

Локация: R208.

Материалы: [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/lecture_03.mp4 запись лекции], [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/seminar_03.mp4 запись семинара], [https://github.com/thecrazymage/DL2_HSE/tree/main/week_03 материалы].

'''Лекция / Семинар 4 (29.09).'''

Тема: Basics of Efficient LLM Training Infrastructure

Аннотация: In this lecture, we will study the basic rules that underpin the infrastructure for efficient large language model training. We will also examine common problems that may arise in this process and explore practical ways to address them.

Лектор и семинарист: [https://habr.com/ru/users/MichaelEk/ Michael Khrushchev]

Локация: лекция - G002, семинар - R208.

Материалы: [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/lecture%2Bseminar_04.mp4 запись лекции и семинара], [https://github.com/thecrazymage/DL2_HSE/tree/main/week_04 материалы].

'''Лекция / Семинар 5 (06.10).'''

Тема: Segmentation and Detection

Аннотация лекции: This week, we'll explore the evolution of object detection and segmentation — from R-CNN to real-time methods like YOLO-World and CLIP-based approaches. We'll examine how U-Net architectures have transcended computer vision to power neural operators and how to use diffusion models for segmentation tasks.

Аннотация семинара: We will implement architectures and train a semantic segmentation model. We will discuss regularization methods for convolutional layers. In addition, we will learn how to use pre-trained models and apply them.

Лектор и семинарист: [https://www.hse.ru/org/persons/956617478/ Alexander Oganov]

Локация: лекция - G002, семинар - R208.

Материалы: [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/lecture_05.mp4 запись лекции], [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/seminar_05.mp4 запись семинара], [https://github.com/thecrazymage/DL2_HSE/tree/main/week_05 материалы].

'''Лекция / Семинар 6 (13.10).'''

Тема: Segmentation and Detection 2

Аннотация лекции и семинара: In this lecture, we will continue in more detail about segmentation and detection.

Лектор: [https://szagoruyko.github.io/ Sergey Zagoruyko]

Семинарист: [https://www.hse.ru/org/persons/401628708/ Eva Neudachina]

Локация: лекция - G002, семинар - R208.

Материалы: [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/lecture_06.mp4 запись лекции], [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/seminar_06.mp4 запись семинара], [https://github.com/thecrazymage/DL2_HSE/tree/main/week_06 материалы].

'''Лекция / Семинар 7 (20.10).'''

Тема: Diffusion models 1

Аннотация лекции и семинара: Tomorrow's lecture and seminar will be devoted to an introduction to diffusion models. Diffusion models are currently the most popular approach to generative modeling due to their high-quality generation and diversity (mode coverage) of the learned distribution. The idea behind diffusion models is to consider the process of gradually transforming data into pure noise and construct its inverse in time, which will transform noise into data. In the lecture and seminar, we will work with noise processes and derive the classic [https://arxiv.org/abs/2006.11239 DDPM] model, which proposes to minimize the KL-divergence between the “true” reverse process that converts noise into data and the denoising process specified by the neural network. In the process, we will see that this procedure is equivalent to training a denoiser neural network that predicts a clean object from a noisy one. In addition, we will interpret the resulting denoising process: in it, each step corresponds to replacing part of the current noisy image with an (increasingly high-quality) prediction of the denoiser.

Лектор и семинарист: [https://www.hse.ru/org/persons/190910999/ Denis Rakitin]

Локация: лекция - G002, семинар - R208.

Материалы: [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/lecture_07.mp4 запись лекции], [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/seminar_07.mp4 запись семинара], [https://github.com/thecrazymage/DL2_HSE/tree/main/week_07 материалы].

'''Лекция / Семинар 8 (03.11).'''

Тема: Diffusion models 2

Аннотация лекции и семинара: In this lecture and seminar, we will continue our exploration of diffusion models. We will introduce the score function and score identity, present classifier and classifier-free guidance, and derive DDIM model.

Лектор и семинарист: [https://www.hse.ru/org/persons/190910999/ Denis Rakitin]

Локация: лекция - R205, семинар - R208.

Материалы: [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/lecture_08.mp4 запись лекции], [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/seminar_08.mp4 запись семинара], [https://github.com/thecrazymage/DL2_HSE/tree/main/week_08 материалы].

'''Лекция / Семинар 9 (10.11).'''

Тема: Diffusion models 3

Аннотация лекции: This lecture presents applied aspects of diffusion models, a class of generative methods that have demonstrated strong performance across images, text, and audio. We will review modern architectures, with a focus on [https://github.com/black-forest-labs/flux FLUX], and unpack the key principles of their design and training.

Лектор: [https://scholar.google.com/citations?user=o6pRm_gAAAAJ&hl=ru Nikita Starodubcev]

Аннотация семинара: The seminar will provide a hands-on, in-depth inspection of [https://github.com/PixArt-alpha/PixArt-alpha PixArt] model. Then ee will attempt further fine-tuning via [https://dreambooth.github.io/ DreamBooth], review common evaluation metrics, and compare different models and configurations.

Семинарист: [https://www.hse.ru/org/persons/401628708/ Eva Neudachina]

Локация: лекция - R205, семинар - R208.

Материалы: [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/lecture_09.mp4 запись лекции], [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/seminar_09.mp4 запись семинара], [https://github.com/thecrazymage/DL2_HSE/tree/main/week_09 материалы].

'''Лекция / Семинар 10 (17.11).'''

Тема: 3D CV

Аннотация лекции и семинара: This lecture and seminar will discuss how diffusion models can be used for 3D computer vision tasks such as generation and reconstruction, and how they relate to methods like NeRF and Gaussian splatting. We will also look at how these models can be applied to problems like novel view synthesis, relighting, and camera pose or depth estimation.

Лектор и семинарист: [https://www.hse.ru/org/persons/885876805/ Mishan Aliev]

Локация: лекция - R205, семинар - R208.

Материалы: [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/lecture_10.mp4 запись лекции], [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/seminar_10.mp4 запись семинара], [https://github.com/thecrazymage/DL2_HSE/tree/main/week_10 материалы].

'''Лекция / Семинар 11 (24.11).'''

Тема: Neural Recommender Systems

Аннотация лекции и семинара: Modern recommender systems also make heavy use of deep learning — both discriminative and generative models. A system has to understand content (tracks, products, videos, etc.), model user behavior (short- and long-term preferences), and predict which content each user is likely to enjoy. At the same time, recommender systems come with their own challenges: cold start for new items and users, huge billion-scale item catalogs with heavy-tailed popularity distributions (popularity bias), and constant distribution drift, where all the underlying distributions keep changing along with the rest of the world. We will discuss: 1) Two-tower neural networks for candidate generation; 2) Sequential recommendation and more modern approaches to generative modeling of users; 3) Semantic IDs and tuning LLMs for recommendation 4) What makes the recommendation domain different from other deep learning domains.

Лектор: [https://www.hse.ru/org/persons/1085421573/ Kirill Khrylchenko]

Семинарист: [https://www.linkedin.com/in/artem-matveev-7b2725255/ Artem Matveev]

Локация: лекция - R205, семинар - R208.

Материалы: [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/lecture_11.mp4 запись лекции], [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/seminar_11.mp4 запись семинара], [https://github.com/thecrazymage/DL2_HSE/tree/main/week_11 материалы].

== Домашние задания ==

{| class="wikitable"
! № !! Домашнее задание !! Ссылка !! Дедлайн (жёсткий)
|-
| 1 || Tensor and DL Libraries || https://github.com/thecrazymage/DL2_HSE/tree/main/homeworks/homework_01 || 30 сентября, 23:59
|-
| 2 || Transformers for Named Entity Recognition || https://github.com/thecrazymage/DL2_HSE/tree/main/homeworks/homework_02 || 14 октября, 23:59
|-
| 3 || Image Segmentation || https://github.com/thecrazymage/DL2_HSE/tree/main/homeworks/homework_03 || 2 ноября, 23:59
|-
| 4 || Diffusion Models || https://github.com/thecrazymage/DL2_HSE/tree/main/homeworks/homework_04 || 18 ноября, 23:59
|-
| 5 || Retrieval || https://github.com/thecrazymage/DL2_HSE/tree/main/homeworks/homework_05 || 10 декабря, 23:59

|}

== Соревнование ==

TBA

Глубинное обучение 2 2025

2025-11-26T17:09:34Z

Anapluslap: add materials for class 11

== Общая информация ==
Курс предназначен для студентов 4 курса ФКН ПМИ (МОП и КНАД).

Занятия проходят '''по понедельникам 14:40-17:40''' (переносы будут сообщаться в чате).

Полезные ссылки:
* Чат с обсуждением: https://t.me/+8dcjl4gHlyEwYzcy
* Репозиторий курса: https://github.com/thecrazymage/DL2_HSE
* Таблица с оценками: https://docs.google.com/spreadsheets/d/18rfZbf7Zsm-xmbiQn_eJVDXXbDfesbHT/edit?usp=sharing&ouid=115132401804687564737&rtpof=true&sd=true
* Anytask: https://anytask.org/course/1219

Формула итоговой оценки (округление арифметическое):
# МОП: Оитог = 0.25 * Осоревнование + 0.75 * ОДЗ,
# КНАД: Оитог = ОДЗ,
где ОДЗ - средняя оценка за практические домашние задания.

== Преподаватели и ассистенты ==

Кому писать, если кажется, что все пропало: [https://t.me/MishanAliev Мишан Алиев]

{| class="wikitable"
|-
! Группа !! Семинарист !! Ассистенты !! Чаты групп !! Инвайт в anytask
|-
| 221 (МОП) || [https://t.me/mightyneighbor Федя Великонивцев] || [https://t.me/vsem_paket Динар Саберов], [https://t.me/annvasileeva Анна Василева] || [https://t.me/+XzlFh0Y8GHs3NTMy МОП 221] || yvZZTIS
|-
| 222 (МОП) || [https://t.me/cocosinca Ева Неудачина] || [https://t.me/mat_os Александр Матосян], [https://t.me/kade1shvili Полина Кадейшвили] || [https://t.me/+PpxiIpF7wLYyMjE6 МОП 222] || W38CfZf
|-
| 223 (МОП) || [https://t.me/tutugarin Иван Ершов] || [https://t.me/Andrew_ut Андрей Уткин], [https://t.me/uzhedevyat Георгий Фатахов] || [https://t.me/+xRZkDO4TjiA1YjQ6 МОП 223] || ph91Jlz
|-
| 224 (МОП) || [https://t.me/Stepyndriy Степан Беляков] || [https://t.me/ponomarchuk_anna Анна Пономарчук], [https://t.me/bogsolntca Татьяна Яковлева] || [https://t.me/+XtcOG352A-IxMmY6 МОП 224] || iPwd342
|-
| КНАД || [https://t.me/burakevi4 Даня Бураков] || [https://t.me/anapluslap Анастасия Лапшина], [https://t.me/irbix7 Иван Галий] || [https://t.me/+JNPUgVxh6ophOGQy КНАД] || zTn4sRP
|}

== Лекции и семинары ==

'''Лекция / Семинар 1 (08.09).'''

Тема: Essentials of GPU, Deep Learning Bottlenecks, and Benchmarking Basics

Аннотация: In this session, we will explore the reasons behind the dominance of GPUs in Deep Learning and examine the common sources of performance bottlenecks in DL code. You will learn how to identify these bottlenecks using profiling tools and apply techniques to optimize and accelerate your code.

Лектор и семинарист: [https://www.hse.ru/org/persons/816100677/ Fedor Velikonivtsev]

Локация: лекция - R208, семинар - D102.

Материалы: [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/lecture_01.mp4 запись лекции], [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/seminar_01.mp4 запись семинара], [https://github.com/thecrazymage/DL2_HSE/tree/main/week_01 материалы].

'''Лекция / Семинар 2 (15.09).'''

Тема: On Transformers and Bitter Lesson

Аннотация: In this talk we’ll dive into the landscape of model architectures in deep learning, with a focus on the world around transformers. We’ll briefly recall what a transformer is, trace the evolution from encoder–decoder to encoder-only and decoder-only models, and touch on the rise of “efficient mixers” such as state space models, linear attention, and beyond. We’ll conclude by reflecting on the role of data and compute. The lecture is inspired in part by [http://incompleteideas.net/IncIdeas/BitterLesson.html The Bitter Lesson] and blends a brain dump with some entertaining insights from recent years of architectural exploration.

Лектор и семинарист: [https://www.hse.ru/org/persons/190912012/ Ivan Rubachev]

Локация: лекция - R206, семинар - R208.

Материалы: [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/lecture%2Bseminar_02.mp4 запись лекции и семинара], [https://github.com/thecrazymage/DL2_HSE/tree/main/week_02 материалы].

'''Лекция / Семинар 3 (22.09).'''

Тема: Modern LLMs essentials

Аннотация лекции: This week, we will discuss LLMs. We will discuss why they are so effective for text generation, how they can be applied to different NLP problems, and the risks they pose. You will learn the details of RLHF, PEFT, and RAG, which make LLMs robust in various cases.

Лектор: [https://www.hse.ru/org/persons/208533329/ Alexander Shabalin]

Локация: онлайн.

Аннотация семинара: In this seminar, we will explore the concept of LLM-based agents and how they extend the capabilities of modern language models. We will discuss function calling as a way to integrate external tools, chain-of-thought reasoning for structured problem solving, and reinforcement learning techniques for training agents.

Семинарист: [http://www.instagram.com/tugarin_vanya Ivan Ershov]

Локация: R208.

Материалы: [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/lecture_03.mp4 запись лекции], [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/seminar_03.mp4 запись семинара], [https://github.com/thecrazymage/DL2_HSE/tree/main/week_03 материалы].

'''Лекция / Семинар 4 (29.09).'''

Тема: Basics of Efficient LLM Training Infrastructure

Аннотация: In this lecture, we will study the basic rules that underpin the infrastructure for efficient large language model training. We will also examine common problems that may arise in this process and explore practical ways to address them.

Лектор и семинарист: [https://habr.com/ru/users/MichaelEk/ Michael Khrushchev]

Локация: лекция - G002, семинар - R208.

Материалы: [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/lecture%2Bseminar_04.mp4 запись лекции и семинара], [https://github.com/thecrazymage/DL2_HSE/tree/main/week_04 материалы].

'''Лекция / Семинар 5 (06.10).'''

Тема: Segmentation and Detection

Аннотация лекции: This week, we'll explore the evolution of object detection and segmentation — from R-CNN to real-time methods like YOLO-World and CLIP-based approaches. We'll examine how U-Net architectures have transcended computer vision to power neural operators and how to use diffusion models for segmentation tasks.

Аннотация семинара: We will implement architectures and train a semantic segmentation model. We will discuss regularization methods for convolutional layers. In addition, we will learn how to use pre-trained models and apply them.

Лектор и семинарист: [https://www.hse.ru/org/persons/956617478/ Alexander Oganov]

Локация: лекция - G002, семинар - R208.

Материалы: [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/lecture_05.mp4 запись лекции], [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/seminar_05.mp4 запись семинара], [https://github.com/thecrazymage/DL2_HSE/tree/main/week_05 материалы].

'''Лекция / Семинар 6 (13.10).'''

Тема: Segmentation and Detection 2

Аннотация лекции и семинара: In this lecture, we will continue in more detail about segmentation and detection.

Лектор: [https://szagoruyko.github.io/ Sergey Zagoruyko]

Семинарист: [https://www.hse.ru/org/persons/401628708/ Eva Neudachina]

Локация: лекция - G002, семинар - R208.

Материалы: [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/lecture_06.mp4 запись лекции], [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/seminar_06.mp4 запись семинара], [https://github.com/thecrazymage/DL2_HSE/tree/main/week_06 материалы].

'''Лекция / Семинар 7 (20.10).'''

Тема: Diffusion models 1

Аннотация лекции и семинара: Tomorrow's lecture and seminar will be devoted to an introduction to diffusion models. Diffusion models are currently the most popular approach to generative modeling due to their high-quality generation and diversity (mode coverage) of the learned distribution. The idea behind diffusion models is to consider the process of gradually transforming data into pure noise and construct its inverse in time, which will transform noise into data. In the lecture and seminar, we will work with noise processes and derive the classic [https://arxiv.org/abs/2006.11239 DDPM] model, which proposes to minimize the KL-divergence between the “true” reverse process that converts noise into data and the denoising process specified by the neural network. In the process, we will see that this procedure is equivalent to training a denoiser neural network that predicts a clean object from a noisy one. In addition, we will interpret the resulting denoising process: in it, each step corresponds to replacing part of the current noisy image with an (increasingly high-quality) prediction of the denoiser.

Лектор и семинарист: [https://www.hse.ru/org/persons/190910999/ Denis Rakitin]

Локация: лекция - G002, семинар - R208.

Материалы: [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/lecture_07.mp4 запись лекции], [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/seminar_07.mp4 запись семинара], [https://github.com/thecrazymage/DL2_HSE/tree/main/week_07 материалы].

'''Лекция / Семинар 8 (03.11).'''

Тема: Diffusion models 2

Аннотация лекции и семинара: In this lecture and seminar, we will continue our exploration of diffusion models. We will introduce the score function and score identity, present classifier and classifier-free guidance, and derive DDIM model.

Лектор и семинарист: [https://www.hse.ru/org/persons/190910999/ Denis Rakitin]

Локация: лекция - R205, семинар - R208.

Материалы: [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/lecture_08.mp4 запись лекции], [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/seminar_08.mp4 запись семинара], [https://github.com/thecrazymage/DL2_HSE/tree/main/week_08 материалы].

'''Лекция / Семинар 9 (10.11).'''

Тема: Diffusion models 3

Аннотация лекции: This lecture presents applied aspects of diffusion models, a class of generative methods that have demonstrated strong performance across images, text, and audio. We will review modern architectures, with a focus on [https://github.com/black-forest-labs/flux FLUX], and unpack the key principles of their design and training.

Лектор: [https://scholar.google.com/citations?user=o6pRm_gAAAAJ&hl=ru Nikita Starodubcev]

Аннотация семинара: The seminar will provide a hands-on, in-depth inspection of [https://github.com/PixArt-alpha/PixArt-alpha PixArt] model. Then ee will attempt further fine-tuning via [https://dreambooth.github.io/ DreamBooth], review common evaluation metrics, and compare different models and configurations.

Семинарист: [https://www.hse.ru/org/persons/401628708/ Eva Neudachina]

Локация: лекция - R205, семинар - R208.

Материалы: [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/lecture_09.mp4 запись лекции], [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/seminar_09.mp4 запись семинара], [https://github.com/thecrazymage/DL2_HSE/tree/main/week_09 материалы].

'''Лекция / Семинар 10 (17.11).'''

Тема: 3D CV

Аннотация лекции и семинара: This lecture and seminar will discuss how diffusion models can be used for 3D computer vision tasks such as generation and reconstruction, and how they relate to methods like NeRF and Gaussian splatting. We will also look at how these models can be applied to problems like novel view synthesis, relighting, and camera pose or depth estimation.

Лектор и семинарист: [https://www.hse.ru/org/persons/885876805/ Mishan Aliev]

Локация: лекция - R205, семинар - R208.

Материалы: [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/lecture_10.mp4 запись лекции], [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/seminar_10.mp4 запись семинара], [https://github.com/thecrazymage/DL2_HSE/tree/main/week_10 материалы].

'''Лекция / Семинар 11 (24.11).'''

Тема: Neural Recommender Systems

Аннотация лекции и семинара: Modern recommender systems also make heavy use of deep learning — both discriminative and generative models. A system has to understand content (tracks, products, videos, etc.), model user behavior (short- and long-term preferences), and predict which content each user is likely to enjoy. At the same time, recommender systems come with their own challenges: cold start for new items and users, huge billion-scale item catalogs with heavy-tailed popularity distributions (popularity bias), and constant distribution drift, where all the underlying distributions keep changing along with the rest of the world. We will discuss: 1) Two-tower neural networks for candidate generation; 2) Sequential recommendation and more modern approaches to generative modeling of users; 3) Semantic IDs and tuning LLMs for recommendation 4) What makes the recommendation domain different from other deep learning domains.

Лектор: [https://www.hse.ru/org/persons/1085421573/ Kirill Khrylchenko]

Семинарист: [https://www.linkedin.com/in/artem-matveev-7b2725255/ Artem Matveev]

Локация: лекция - R205, семинар - R208.

Материалы: [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/lecture_11.mp4 запись лекции], [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/seminar_11.mp4 запись семинара], [https://github.com/thecrazymage/DL2_HSE/tree/main/week_11 материалы].

== Домашние задания ==

{| class="wikitable"
! № !! Домашнее задание !! Ссылка !! Дедлайн (жёсткий)
|-
| 1 || Tensor and DL Libraries || https://github.com/thecrazymage/DL2_HSE/tree/main/homeworks/homework_01 || 30 сентября, 23:59
|-
| 2 || Transformers for Named Entity Recognition || https://github.com/thecrazymage/DL2_HSE/tree/main/homeworks/homework_02 || 14 октября, 23:59
|-
| 3 || Image Segmentation || https://github.com/thecrazymage/DL2_HSE/tree/main/homeworks/homework_03 || 2 ноября, 23:59
|-
| 4 || Diffusion Models || https://github.com/thecrazymage/DL2_HSE/tree/main/homeworks/homework_04 || 18 ноября, 23:59

|}

== Соревнование ==

TBA

Глубинное обучение 2 2025

2025-11-24T10:12:26Z

Anapluslap: add class 11 announcement

== Общая информация ==
Курс предназначен для студентов 4 курса ФКН ПМИ (МОП и КНАД).

Занятия проходят '''по понедельникам 14:40-17:40''' (переносы будут сообщаться в чате).

Полезные ссылки:
* Чат с обсуждением: https://t.me/+8dcjl4gHlyEwYzcy
* Репозиторий курса: https://github.com/thecrazymage/DL2_HSE
* Таблица с оценками: https://docs.google.com/spreadsheets/d/18rfZbf7Zsm-xmbiQn_eJVDXXbDfesbHT/edit?usp=sharing&ouid=115132401804687564737&rtpof=true&sd=true
* Anytask: https://anytask.org/course/1219

Формула итоговой оценки (округление арифметическое):
# МОП: Оитог = 0.25 * Осоревнование + 0.75 * ОДЗ,
# КНАД: Оитог = ОДЗ,
где ОДЗ - средняя оценка за практические домашние задания.

== Преподаватели и ассистенты ==

Кому писать, если кажется, что все пропало: [https://t.me/MishanAliev Мишан Алиев]

{| class="wikitable"
|-
! Группа !! Семинарист !! Ассистенты !! Чаты групп !! Инвайт в anytask
|-
| 221 (МОП) || [https://t.me/mightyneighbor Федя Великонивцев] || [https://t.me/vsem_paket Динар Саберов], [https://t.me/annvasileeva Анна Василева] || [https://t.me/+XzlFh0Y8GHs3NTMy МОП 221] || yvZZTIS
|-
| 222 (МОП) || [https://t.me/cocosinca Ева Неудачина] || [https://t.me/mat_os Александр Матосян], [https://t.me/kade1shvili Полина Кадейшвили] || [https://t.me/+PpxiIpF7wLYyMjE6 МОП 222] || W38CfZf
|-
| 223 (МОП) || [https://t.me/tutugarin Иван Ершов] || [https://t.me/Andrew_ut Андрей Уткин], [https://t.me/uzhedevyat Георгий Фатахов] || [https://t.me/+xRZkDO4TjiA1YjQ6 МОП 223] || ph91Jlz
|-
| 224 (МОП) || [https://t.me/Stepyndriy Степан Беляков] || [https://t.me/ponomarchuk_anna Анна Пономарчук], [https://t.me/bogsolntca Татьяна Яковлева] || [https://t.me/+XtcOG352A-IxMmY6 МОП 224] || iPwd342
|-
| КНАД || [https://t.me/burakevi4 Даня Бураков] || [https://t.me/anapluslap Анастасия Лапшина], [https://t.me/irbix7 Иван Галий] || [https://t.me/+JNPUgVxh6ophOGQy КНАД] || zTn4sRP
|}

== Лекции и семинары ==

'''Лекция / Семинар 1 (08.09).'''

Тема: Essentials of GPU, Deep Learning Bottlenecks, and Benchmarking Basics

Аннотация: In this session, we will explore the reasons behind the dominance of GPUs in Deep Learning and examine the common sources of performance bottlenecks in DL code. You will learn how to identify these bottlenecks using profiling tools and apply techniques to optimize and accelerate your code.

Лектор и семинарист: [https://www.hse.ru/org/persons/816100677/ Fedor Velikonivtsev]

Локация: лекция - R208, семинар - D102.

Материалы: [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/lecture_01.mp4 запись лекции], [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/seminar_01.mp4 запись семинара], [https://github.com/thecrazymage/DL2_HSE/tree/main/week_01 материалы].

'''Лекция / Семинар 2 (15.09).'''

Тема: On Transformers and Bitter Lesson

Аннотация: In this talk we’ll dive into the landscape of model architectures in deep learning, with a focus on the world around transformers. We’ll briefly recall what a transformer is, trace the evolution from encoder–decoder to encoder-only and decoder-only models, and touch on the rise of “efficient mixers” such as state space models, linear attention, and beyond. We’ll conclude by reflecting on the role of data and compute. The lecture is inspired in part by [http://incompleteideas.net/IncIdeas/BitterLesson.html The Bitter Lesson] and blends a brain dump with some entertaining insights from recent years of architectural exploration.

Лектор и семинарист: [https://www.hse.ru/org/persons/190912012/ Ivan Rubachev]

Локация: лекция - R206, семинар - R208.

Материалы: [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/lecture%2Bseminar_02.mp4 запись лекции и семинара], [https://github.com/thecrazymage/DL2_HSE/tree/main/week_02 материалы].

'''Лекция / Семинар 3 (22.09).'''

Тема: Modern LLMs essentials

Аннотация лекции: This week, we will discuss LLMs. We will discuss why they are so effective for text generation, how they can be applied to different NLP problems, and the risks they pose. You will learn the details of RLHF, PEFT, and RAG, which make LLMs robust in various cases.

Лектор: [https://www.hse.ru/org/persons/208533329/ Alexander Shabalin]

Локация: онлайн.

Аннотация семинара: In this seminar, we will explore the concept of LLM-based agents and how they extend the capabilities of modern language models. We will discuss function calling as a way to integrate external tools, chain-of-thought reasoning for structured problem solving, and reinforcement learning techniques for training agents.

Семинарист: [http://www.instagram.com/tugarin_vanya Ivan Ershov]

Локация: R208.

Материалы: [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/lecture_03.mp4 запись лекции], [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/seminar_03.mp4 запись семинара], [https://github.com/thecrazymage/DL2_HSE/tree/main/week_03 материалы].

'''Лекция / Семинар 4 (29.09).'''

Тема: Basics of Efficient LLM Training Infrastructure

Аннотация: In this lecture, we will study the basic rules that underpin the infrastructure for efficient large language model training. We will also examine common problems that may arise in this process and explore practical ways to address them.

Лектор и семинарист: [https://habr.com/ru/users/MichaelEk/ Michael Khrushchev]

Локация: лекция - G002, семинар - R208.

Материалы: [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/lecture%2Bseminar_04.mp4 запись лекции и семинара], [https://github.com/thecrazymage/DL2_HSE/tree/main/week_04 материалы].

'''Лекция / Семинар 5 (06.10).'''

Тема: Segmentation and Detection

Аннотация лекции: This week, we'll explore the evolution of object detection and segmentation — from R-CNN to real-time methods like YOLO-World and CLIP-based approaches. We'll examine how U-Net architectures have transcended computer vision to power neural operators and how to use diffusion models for segmentation tasks.

Аннотация семинара: We will implement architectures and train a semantic segmentation model. We will discuss regularization methods for convolutional layers. In addition, we will learn how to use pre-trained models and apply them.

Лектор и семинарист: [https://www.hse.ru/org/persons/956617478/ Alexander Oganov]

Локация: лекция - G002, семинар - R208.

Материалы: [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/lecture_05.mp4 запись лекции], [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/seminar_05.mp4 запись семинара], [https://github.com/thecrazymage/DL2_HSE/tree/main/week_05 материалы].

'''Лекция / Семинар 6 (13.10).'''

Тема: Segmentation and Detection 2

Аннотация лекции и семинара: In this lecture, we will continue in more detail about segmentation and detection.

Лектор: [https://szagoruyko.github.io/ Sergey Zagoruyko]

Семинарист: [https://www.hse.ru/org/persons/401628708/ Eva Neudachina]

Локация: лекция - G002, семинар - R208.

Материалы: [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/lecture_06.mp4 запись лекции], [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/seminar_06.mp4 запись семинара], [https://github.com/thecrazymage/DL2_HSE/tree/main/week_06 материалы].

'''Лекция / Семинар 7 (20.10).'''

Тема: Diffusion models 1

Аннотация лекции и семинара: Tomorrow's lecture and seminar will be devoted to an introduction to diffusion models. Diffusion models are currently the most popular approach to generative modeling due to their high-quality generation and diversity (mode coverage) of the learned distribution. The idea behind diffusion models is to consider the process of gradually transforming data into pure noise and construct its inverse in time, which will transform noise into data. In the lecture and seminar, we will work with noise processes and derive the classic [https://arxiv.org/abs/2006.11239 DDPM] model, which proposes to minimize the KL-divergence between the “true” reverse process that converts noise into data and the denoising process specified by the neural network. In the process, we will see that this procedure is equivalent to training a denoiser neural network that predicts a clean object from a noisy one. In addition, we will interpret the resulting denoising process: in it, each step corresponds to replacing part of the current noisy image with an (increasingly high-quality) prediction of the denoiser.

Лектор и семинарист: [https://www.hse.ru/org/persons/190910999/ Denis Rakitin]

Локация: лекция - G002, семинар - R208.

Материалы: [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/lecture_07.mp4 запись лекции], [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/seminar_07.mp4 запись семинара], [https://github.com/thecrazymage/DL2_HSE/tree/main/week_07 материалы].

'''Лекция / Семинар 8 (03.11).'''

Тема: Diffusion models 2

Аннотация лекции и семинара: In this lecture and seminar, we will continue our exploration of diffusion models. We will introduce the score function and score identity, present classifier and classifier-free guidance, and derive DDIM model.

Лектор и семинарист: [https://www.hse.ru/org/persons/190910999/ Denis Rakitin]

Локация: лекция - R205, семинар - R208.

Материалы: [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/lecture_08.mp4 запись лекции], [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/seminar_08.mp4 запись семинара], [https://github.com/thecrazymage/DL2_HSE/tree/main/week_08 материалы].

'''Лекция / Семинар 9 (10.11).'''

Тема: Diffusion models 3

Аннотация лекции: This lecture presents applied aspects of diffusion models, a class of generative methods that have demonstrated strong performance across images, text, and audio. We will review modern architectures, with a focus on [https://github.com/black-forest-labs/flux FLUX], and unpack the key principles of their design and training.

Лектор: [https://scholar.google.com/citations?user=o6pRm_gAAAAJ&hl=ru Nikita Starodubcev]

Аннотация семинара: The seminar will provide a hands-on, in-depth inspection of [https://github.com/PixArt-alpha/PixArt-alpha PixArt] model. Then ee will attempt further fine-tuning via [https://dreambooth.github.io/ DreamBooth], review common evaluation metrics, and compare different models and configurations.

Семинарист: [https://www.hse.ru/org/persons/401628708/ Eva Neudachina]

Локация: лекция - R205, семинар - R208.

Материалы: [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/lecture_09.mp4 запись лекции], [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/seminar_09.mp4 запись семинара], [https://github.com/thecrazymage/DL2_HSE/tree/main/week_09 материалы].

'''Лекция / Семинар 10 (17.11).'''

Тема: 3D CV

Аннотация лекции и семинара: This lecture and seminar will discuss how diffusion models can be used for 3D computer vision tasks such as generation and reconstruction, and how they relate to methods like NeRF and Gaussian splatting. We will also look at how these models can be applied to problems like novel view synthesis, relighting, and camera pose or depth estimation.

Лектор и семинарист: [https://www.hse.ru/org/persons/885876805/ Mishan Aliev]

Локация: лекция - R205, семинар - R208.

Материалы: [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/lecture_10.mp4 запись лекции], [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/seminar_10.mp4 запись семинара], [https://github.com/thecrazymage/DL2_HSE/tree/main/week_10 материалы].

'''Лекция / Семинар 11 (24.11).'''

Тема: Neural Recommender Systems

Аннотация лекции и семинара: Modern recommender systems also make heavy use of deep learning — both discriminative and generative models. A system has to understand content (tracks, products, videos, etc.), model user behavior (short- and long-term preferences), and predict which content each user is likely to enjoy. At the same time, recommender systems come with their own challenges: cold start for new items and users, huge billion-scale item catalogs with heavy-tailed popularity distributions (popularity bias), and constant distribution drift, where all the underlying distributions keep changing along with the rest of the world. We will discuss: 1) Two-tower neural networks for candidate generation; 2) Sequential recommendation and more modern approaches to generative modeling of users; 3) Semantic IDs and tuning LLMs for recommendation 4) What makes the recommendation domain different from other deep learning domains.

Лектор: [https://www.hse.ru/org/persons/1085421573/ Kirill Khrylchenko]

Семинарист: [https://www.linkedin.com/in/artem-matveev-7b2725255/ Artem Matveev]

Локация: лекция - R205, семинар - R208.

Материалы: tba.

== Домашние задания ==

{| class="wikitable"
! № !! Домашнее задание !! Ссылка !! Дедлайн (жёсткий)
|-
| 1 || Tensor and DL Libraries || https://github.com/thecrazymage/DL2_HSE/tree/main/homeworks/homework_01 || 30 сентября, 23:59
|-
| 2 || Transformers for Named Entity Recognition || https://github.com/thecrazymage/DL2_HSE/tree/main/homeworks/homework_02 || 14 октября, 23:59
|-
| 3 || Image Segmentation || https://github.com/thecrazymage/DL2_HSE/tree/main/homeworks/homework_03 || 2 ноября, 23:59
|-
| 4 || Diffusion Models || https://github.com/thecrazymage/DL2_HSE/tree/main/homeworks/homework_04 || 18 ноября, 23:59

|}

== Соревнование ==

TBA

Глубинное обучение 2 2025

2025-11-18T18:02:06Z

Anapluslap: add class 10 materials

== Общая информация ==
Курс предназначен для студентов 4 курса ФКН ПМИ (МОП и КНАД).

Занятия проходят '''по понедельникам 14:40-17:40''' (переносы будут сообщаться в чате).

Полезные ссылки:
* Чат с обсуждением: https://t.me/+8dcjl4gHlyEwYzcy
* Репозиторий курса: https://github.com/thecrazymage/DL2_HSE
* Таблица с оценками: https://docs.google.com/spreadsheets/d/18rfZbf7Zsm-xmbiQn_eJVDXXbDfesbHT/edit?usp=sharing&ouid=115132401804687564737&rtpof=true&sd=true
* Anytask: https://anytask.org/course/1219

Формула итоговой оценки (округление арифметическое):
# МОП: Оитог = 0.25 * Осоревнование + 0.75 * ОДЗ,
# КНАД: Оитог = ОДЗ,
где ОДЗ - средняя оценка за практические домашние задания.

== Преподаватели и ассистенты ==

Кому писать, если кажется, что все пропало: [https://t.me/MishanAliev Мишан Алиев]

{| class="wikitable"
|-
! Группа !! Семинарист !! Ассистенты !! Чаты групп !! Инвайт в anytask
|-
| 221 (МОП) || [https://t.me/mightyneighbor Федя Великонивцев] || [https://t.me/vsem_paket Динар Саберов], [https://t.me/annvasileeva Анна Василева] || [https://t.me/+XzlFh0Y8GHs3NTMy МОП 221] || yvZZTIS
|-
| 222 (МОП) || [https://t.me/cocosinca Ева Неудачина] || [https://t.me/mat_os Александр Матосян], [https://t.me/kade1shvili Полина Кадейшвили] || [https://t.me/+PpxiIpF7wLYyMjE6 МОП 222] || W38CfZf
|-
| 223 (МОП) || [https://t.me/tutugarin Иван Ершов] || [https://t.me/Andrew_ut Андрей Уткин], [https://t.me/uzhedevyat Георгий Фатахов] || [https://t.me/+xRZkDO4TjiA1YjQ6 МОП 223] || ph91Jlz
|-
| 224 (МОП) || [https://t.me/Stepyndriy Степан Беляков] || [https://t.me/ponomarchuk_anna Анна Пономарчук], [https://t.me/bogsolntca Татьяна Яковлева] || [https://t.me/+XtcOG352A-IxMmY6 МОП 224] || iPwd342
|-
| КНАД || [https://t.me/burakevi4 Даня Бураков] || [https://t.me/anapluslap Анастасия Лапшина], [https://t.me/irbix7 Иван Галий] || [https://t.me/+JNPUgVxh6ophOGQy КНАД] || zTn4sRP
|}

== Лекции и семинары ==

'''Лекция / Семинар 1 (08.09).'''

Тема: Essentials of GPU, Deep Learning Bottlenecks, and Benchmarking Basics

Аннотация: In this session, we will explore the reasons behind the dominance of GPUs in Deep Learning and examine the common sources of performance bottlenecks in DL code. You will learn how to identify these bottlenecks using profiling tools and apply techniques to optimize and accelerate your code.

Лектор и семинарист: [https://www.hse.ru/org/persons/816100677/ Fedor Velikonivtsev]

Локация: лекция - R208, семинар - D102.

Материалы: [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/lecture_01.mp4 запись лекции], [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/seminar_01.mp4 запись семинара], [https://github.com/thecrazymage/DL2_HSE/tree/main/week_01 материалы].

'''Лекция / Семинар 2 (15.09).'''

Тема: On Transformers and Bitter Lesson

Аннотация: In this talk we’ll dive into the landscape of model architectures in deep learning, with a focus on the world around transformers. We’ll briefly recall what a transformer is, trace the evolution from encoder–decoder to encoder-only and decoder-only models, and touch on the rise of “efficient mixers” such as state space models, linear attention, and beyond. We’ll conclude by reflecting on the role of data and compute. The lecture is inspired in part by [http://incompleteideas.net/IncIdeas/BitterLesson.html The Bitter Lesson] and blends a brain dump with some entertaining insights from recent years of architectural exploration.

Лектор и семинарист: [https://www.hse.ru/org/persons/190912012/ Ivan Rubachev]

Локация: лекция - R206, семинар - R208.

Материалы: [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/lecture%2Bseminar_02.mp4 запись лекции и семинара], [https://github.com/thecrazymage/DL2_HSE/tree/main/week_02 материалы].

'''Лекция / Семинар 3 (22.09).'''

Тема: Modern LLMs essentials

Аннотация лекции: This week, we will discuss LLMs. We will discuss why they are so effective for text generation, how they can be applied to different NLP problems, and the risks they pose. You will learn the details of RLHF, PEFT, and RAG, which make LLMs robust in various cases.

Лектор: [https://www.hse.ru/org/persons/208533329/ Alexander Shabalin]

Локация: онлайн.

Аннотация семинара: In this seminar, we will explore the concept of LLM-based agents and how they extend the capabilities of modern language models. We will discuss function calling as a way to integrate external tools, chain-of-thought reasoning for structured problem solving, and reinforcement learning techniques for training agents.

Семинарист: [http://www.instagram.com/tugarin_vanya Ivan Ershov]

Локация: R208.

Материалы: [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/lecture_03.mp4 запись лекции], [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/seminar_03.mp4 запись семинара], [https://github.com/thecrazymage/DL2_HSE/tree/main/week_03 материалы].

'''Лекция / Семинар 4 (29.09).'''

Тема: Basics of Efficient LLM Training Infrastructure

Аннотация: In this lecture, we will study the basic rules that underpin the infrastructure for efficient large language model training. We will also examine common problems that may arise in this process and explore practical ways to address them.

Лектор и семинарист: [https://habr.com/ru/users/MichaelEk/ Michael Khrushchev]

Локация: лекция - G002, семинар - R208.

Материалы: [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/lecture%2Bseminar_04.mp4 запись лекции и семинара], [https://github.com/thecrazymage/DL2_HSE/tree/main/week_04 материалы].

'''Лекция / Семинар 5 (06.10).'''

Тема: Segmentation and Detection

Аннотация лекции: This week, we'll explore the evolution of object detection and segmentation — from R-CNN to real-time methods like YOLO-World and CLIP-based approaches. We'll examine how U-Net architectures have transcended computer vision to power neural operators and how to use diffusion models for segmentation tasks.

Аннотация семинара: We will implement architectures and train a semantic segmentation model. We will discuss regularization methods for convolutional layers. In addition, we will learn how to use pre-trained models and apply them.

Лектор и семинарист: [https://www.hse.ru/org/persons/956617478/ Alexander Oganov]

Локация: лекция - G002, семинар - R208.

Материалы: [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/lecture_05.mp4 запись лекции], [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/seminar_05.mp4 запись семинара], [https://github.com/thecrazymage/DL2_HSE/tree/main/week_05 материалы].

'''Лекция / Семинар 6 (13.10).'''

Тема: Segmentation and Detection 2

Аннотация лекции и семинара: In this lecture, we will continue in more detail about segmentation and detection.

Лектор: [https://szagoruyko.github.io/ Sergey Zagoruyko]

Семинарист: [https://www.hse.ru/org/persons/401628708/ Eva Neudachina]

Локация: лекция - G002, семинар - R208.

Материалы: [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/lecture_06.mp4 запись лекции], [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/seminar_06.mp4 запись семинара], [https://github.com/thecrazymage/DL2_HSE/tree/main/week_06 материалы].

'''Лекция / Семинар 7 (20.10).'''

Тема: Diffusion models 1

Аннотация лекции и семинара: Tomorrow's lecture and seminar will be devoted to an introduction to diffusion models. Diffusion models are currently the most popular approach to generative modeling due to their high-quality generation and diversity (mode coverage) of the learned distribution. The idea behind diffusion models is to consider the process of gradually transforming data into pure noise and construct its inverse in time, which will transform noise into data. In the lecture and seminar, we will work with noise processes and derive the classic [https://arxiv.org/abs/2006.11239 DDPM] model, which proposes to minimize the KL-divergence between the “true” reverse process that converts noise into data and the denoising process specified by the neural network. In the process, we will see that this procedure is equivalent to training a denoiser neural network that predicts a clean object from a noisy one. In addition, we will interpret the resulting denoising process: in it, each step corresponds to replacing part of the current noisy image with an (increasingly high-quality) prediction of the denoiser.

Лектор и семинарист: [https://www.hse.ru/org/persons/190910999/ Denis Rakitin]

Локация: лекция - G002, семинар - R208.

Материалы: [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/lecture_07.mp4 запись лекции], [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/seminar_07.mp4 запись семинара], [https://github.com/thecrazymage/DL2_HSE/tree/main/week_07 материалы].

'''Лекция / Семинар 8 (03.11).'''

Тема: Diffusion models 2

Аннотация лекции и семинара: In this lecture and seminar, we will continue our exploration of diffusion models. We will introduce the score function and score identity, present classifier and classifier-free guidance, and derive DDIM model.

Лектор и семинарист: [https://www.hse.ru/org/persons/190910999/ Denis Rakitin]

Локация: лекция - R205, семинар - R208.

Материалы: [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/lecture_08.mp4 запись лекции], [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/seminar_08.mp4 запись семинара], [https://github.com/thecrazymage/DL2_HSE/tree/main/week_08 материалы].

'''Лекция / Семинар 9 (10.11).'''

Тема: Diffusion models 3

Аннотация лекции: This lecture presents applied aspects of diffusion models, a class of generative methods that have demonstrated strong performance across images, text, and audio. We will review modern architectures, with a focus on [https://github.com/black-forest-labs/flux FLUX], and unpack the key principles of their design and training.

Лектор: [https://scholar.google.com/citations?user=o6pRm_gAAAAJ&hl=ru Nikita Starodubcev]

Аннотация семинара: The seminar will provide a hands-on, in-depth inspection of [https://github.com/PixArt-alpha/PixArt-alpha PixArt] model. Then ee will attempt further fine-tuning via [https://dreambooth.github.io/ DreamBooth], review common evaluation metrics, and compare different models and configurations.

Семинарист: [https://www.hse.ru/org/persons/401628708/ Eva Neudachina]

Локация: лекция - R205, семинар - R208.

Материалы: [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/lecture_09.mp4 запись лекции], [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/seminar_09.mp4 запись семинара], [https://github.com/thecrazymage/DL2_HSE/tree/main/week_09 материалы].

'''Лекция / Семинар 10 (17.11).'''

Тема: 3D CV

Аннотация лекции и семинара: This lecture and seminar will discuss how diffusion models can be used for 3D computer vision tasks such as generation and reconstruction, and how they relate to methods like NeRF and Gaussian splatting. We will also look at how these models can be applied to problems like novel view synthesis, relighting, and camera pose or depth estimation.

Лектор и семинарист: [https://www.hse.ru/org/persons/885876805/ Mishan Aliev]

Локация: лекция - R205, семинар - R208.

Материалы: [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/lecture_10.mp4 запись лекции], [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/seminar_10.mp4 запись семинара], [https://github.com/thecrazymage/DL2_HSE/tree/main/week_10 материалы].

== Домашние задания ==

{| class="wikitable"
! № !! Домашнее задание !! Ссылка !! Дедлайн (жёсткий)
|-
| 1 || Tensor and DL Libraries || https://github.com/thecrazymage/DL2_HSE/tree/main/homeworks/homework_01 || 30 сентября, 23:59
|-
| 2 || Transformers for Named Entity Recognition || https://github.com/thecrazymage/DL2_HSE/tree/main/homeworks/homework_02 || 14 октября, 23:59
|-
| 3 || Image Segmentation || https://github.com/thecrazymage/DL2_HSE/tree/main/homeworks/homework_03 || 2 ноября, 23:59
|-
| 4 || Diffusion Models || https://github.com/thecrazymage/DL2_HSE/tree/main/homeworks/homework_04 || 18 ноября, 23:59

|}

== Соревнование ==

TBA

Глубинное обучение 2 2025

2025-11-15T11:30:33Z

Anapluslap: add class 10 anouncement

== Общая информация ==
Курс предназначен для студентов 4 курса ФКН ПМИ (МОП и КНАД).

Занятия проходят '''по понедельникам 14:40-17:40''' (переносы будут сообщаться в чате).

Полезные ссылки:
* Чат с обсуждением: https://t.me/+8dcjl4gHlyEwYzcy
* Репозиторий курса: https://github.com/thecrazymage/DL2_HSE
* Таблица с оценками: https://docs.google.com/spreadsheets/d/18rfZbf7Zsm-xmbiQn_eJVDXXbDfesbHT/edit?usp=sharing&ouid=115132401804687564737&rtpof=true&sd=true
* Anytask: https://anytask.org/course/1219

Формула итоговой оценки (округление арифметическое):
# МОП: Оитог = 0.25 * Осоревнование + 0.75 * ОДЗ,
# КНАД: Оитог = ОДЗ,
где ОДЗ - средняя оценка за практические домашние задания.

== Преподаватели и ассистенты ==

Кому писать, если кажется, что все пропало: [https://t.me/MishanAliev Мишан Алиев]

{| class="wikitable"
|-
! Группа !! Семинарист !! Ассистенты !! Чаты групп !! Инвайт в anytask
|-
| 221 (МОП) || [https://t.me/mightyneighbor Федя Великонивцев] || [https://t.me/vsem_paket Динар Саберов], [https://t.me/annvasileeva Анна Василева] || [https://t.me/+XzlFh0Y8GHs3NTMy МОП 221] || yvZZTIS
|-
| 222 (МОП) || [https://t.me/cocosinca Ева Неудачина] || [https://t.me/mat_os Александр Матосян], [https://t.me/kade1shvili Полина Кадейшвили] || [https://t.me/+PpxiIpF7wLYyMjE6 МОП 222] || W38CfZf
|-
| 223 (МОП) || [https://t.me/tutugarin Иван Ершов] || [https://t.me/Andrew_ut Андрей Уткин], [https://t.me/uzhedevyat Георгий Фатахов] || [https://t.me/+xRZkDO4TjiA1YjQ6 МОП 223] || ph91Jlz
|-
| 224 (МОП) || [https://t.me/Stepyndriy Степан Беляков] || [https://t.me/ponomarchuk_anna Анна Пономарчук], [https://t.me/bogsolntca Татьяна Яковлева] || [https://t.me/+XtcOG352A-IxMmY6 МОП 224] || iPwd342
|-
| КНАД || [https://t.me/burakevi4 Даня Бураков] || [https://t.me/anapluslap Анастасия Лапшина], [https://t.me/irbix7 Иван Галий] || [https://t.me/+JNPUgVxh6ophOGQy КНАД] || zTn4sRP
|}

== Лекции и семинары ==

'''Лекция / Семинар 1 (08.09).'''

Тема: Essentials of GPU, Deep Learning Bottlenecks, and Benchmarking Basics

Аннотация: In this session, we will explore the reasons behind the dominance of GPUs in Deep Learning and examine the common sources of performance bottlenecks in DL code. You will learn how to identify these bottlenecks using profiling tools and apply techniques to optimize and accelerate your code.

Лектор и семинарист: [https://www.hse.ru/org/persons/816100677/ Fedor Velikonivtsev]

Локация: лекция - R208, семинар - D102.

Материалы: [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/lecture_01.mp4 запись лекции], [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/seminar_01.mp4 запись семинара], [https://github.com/thecrazymage/DL2_HSE/tree/main/week_01 материалы].

'''Лекция / Семинар 2 (15.09).'''

Тема: On Transformers and Bitter Lesson

Аннотация: In this talk we’ll dive into the landscape of model architectures in deep learning, with a focus on the world around transformers. We’ll briefly recall what a transformer is, trace the evolution from encoder–decoder to encoder-only and decoder-only models, and touch on the rise of “efficient mixers” such as state space models, linear attention, and beyond. We’ll conclude by reflecting on the role of data and compute. The lecture is inspired in part by [http://incompleteideas.net/IncIdeas/BitterLesson.html The Bitter Lesson] and blends a brain dump with some entertaining insights from recent years of architectural exploration.

Лектор и семинарист: [https://www.hse.ru/org/persons/190912012/ Ivan Rubachev]

Локация: лекция - R206, семинар - R208.

Материалы: [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/lecture%2Bseminar_02.mp4 запись лекции и семинара], [https://github.com/thecrazymage/DL2_HSE/tree/main/week_02 материалы].

'''Лекция / Семинар 3 (22.09).'''

Тема: Modern LLMs essentials

Аннотация лекции: This week, we will discuss LLMs. We will discuss why they are so effective for text generation, how they can be applied to different NLP problems, and the risks they pose. You will learn the details of RLHF, PEFT, and RAG, which make LLMs robust in various cases.

Лектор: [https://www.hse.ru/org/persons/208533329/ Alexander Shabalin]

Локация: онлайн.

Аннотация семинара: In this seminar, we will explore the concept of LLM-based agents and how they extend the capabilities of modern language models. We will discuss function calling as a way to integrate external tools, chain-of-thought reasoning for structured problem solving, and reinforcement learning techniques for training agents.

Семинарист: [http://www.instagram.com/tugarin_vanya Ivan Ershov]

Локация: R208.

Материалы: [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/lecture_03.mp4 запись лекции], [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/seminar_03.mp4 запись семинара], [https://github.com/thecrazymage/DL2_HSE/tree/main/week_03 материалы].

'''Лекция / Семинар 4 (29.09).'''

Тема: Basics of Efficient LLM Training Infrastructure

Аннотация: In this lecture, we will study the basic rules that underpin the infrastructure for efficient large language model training. We will also examine common problems that may arise in this process and explore practical ways to address them.

Лектор и семинарист: [https://habr.com/ru/users/MichaelEk/ Michael Khrushchev]

Локация: лекция - G002, семинар - R208.

Материалы: [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/lecture%2Bseminar_04.mp4 запись лекции и семинара], [https://github.com/thecrazymage/DL2_HSE/tree/main/week_04 материалы].

'''Лекция / Семинар 5 (06.10).'''

Тема: Segmentation and Detection

Аннотация лекции: This week, we'll explore the evolution of object detection and segmentation — from R-CNN to real-time methods like YOLO-World and CLIP-based approaches. We'll examine how U-Net architectures have transcended computer vision to power neural operators and how to use diffusion models for segmentation tasks.

Аннотация семинара: We will implement architectures and train a semantic segmentation model. We will discuss regularization methods for convolutional layers. In addition, we will learn how to use pre-trained models and apply them.

Лектор и семинарист: [https://www.hse.ru/org/persons/956617478/ Alexander Oganov]

Локация: лекция - G002, семинар - R208.

Материалы: [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/lecture_05.mp4 запись лекции], [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/seminar_05.mp4 запись семинара], [https://github.com/thecrazymage/DL2_HSE/tree/main/week_05 материалы].

'''Лекция / Семинар 6 (13.10).'''

Тема: Segmentation and Detection 2

Аннотация лекции и семинара: In this lecture, we will continue in more detail about segmentation and detection.

Лектор: [https://szagoruyko.github.io/ Sergey Zagoruyko]

Семинарист: [https://www.hse.ru/org/persons/401628708/ Eva Neudachina]

Локация: лекция - G002, семинар - R208.

Материалы: [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/lecture_06.mp4 запись лекции], [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/seminar_06.mp4 запись семинара], [https://github.com/thecrazymage/DL2_HSE/tree/main/week_06 материалы].

'''Лекция / Семинар 7 (20.10).'''

Тема: Diffusion models 1

Аннотация лекции и семинара: Tomorrow's lecture and seminar will be devoted to an introduction to diffusion models. Diffusion models are currently the most popular approach to generative modeling due to their high-quality generation and diversity (mode coverage) of the learned distribution. The idea behind diffusion models is to consider the process of gradually transforming data into pure noise and construct its inverse in time, which will transform noise into data. In the lecture and seminar, we will work with noise processes and derive the classic [https://arxiv.org/abs/2006.11239 DDPM] model, which proposes to minimize the KL-divergence between the “true” reverse process that converts noise into data and the denoising process specified by the neural network. In the process, we will see that this procedure is equivalent to training a denoiser neural network that predicts a clean object from a noisy one. In addition, we will interpret the resulting denoising process: in it, each step corresponds to replacing part of the current noisy image with an (increasingly high-quality) prediction of the denoiser.

Лектор и семинарист: [https://www.hse.ru/org/persons/190910999/ Denis Rakitin]

Локация: лекция - G002, семинар - R208.

Материалы: [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/lecture_07.mp4 запись лекции], [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/seminar_07.mp4 запись семинара], [https://github.com/thecrazymage/DL2_HSE/tree/main/week_07 материалы].

'''Лекция / Семинар 8 (03.11).'''

Тема: Diffusion models 2

Аннотация лекции и семинара: In this lecture and seminar, we will continue our exploration of diffusion models. We will introduce the score function and score identity, present classifier and classifier-free guidance, and derive DDIM model.

Лектор и семинарист: [https://www.hse.ru/org/persons/190910999/ Denis Rakitin]

Локация: лекция - R205, семинар - R208.

Материалы: [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/lecture_08.mp4 запись лекции], [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/seminar_08.mp4 запись семинара], [https://github.com/thecrazymage/DL2_HSE/tree/main/week_08 материалы].

'''Лекция / Семинар 9 (10.11).'''

Тема: Diffusion models 3

Аннотация лекции: This lecture presents applied aspects of diffusion models, a class of generative methods that have demonstrated strong performance across images, text, and audio. We will review modern architectures, with a focus on [https://github.com/black-forest-labs/flux FLUX], and unpack the key principles of their design and training.

Лектор: [https://scholar.google.com/citations?user=o6pRm_gAAAAJ&hl=ru Nikita Starodubcev]

Аннотация семинара: The seminar will provide a hands-on, in-depth inspection of [https://github.com/PixArt-alpha/PixArt-alpha PixArt] model. Then ee will attempt further fine-tuning via [https://dreambooth.github.io/ DreamBooth], review common evaluation metrics, and compare different models and configurations.

Семинарист: [https://www.hse.ru/org/persons/401628708/ Eva Neudachina]

Локация: лекция - R205, семинар - R208.

Материалы: [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/lecture_09.mp4 запись лекции], [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/seminar_09.mp4 запись семинара], [https://github.com/thecrazymage/DL2_HSE/tree/main/week_09 материалы].

'''Лекция / Семинар 10 (17.11).'''

Тема: 3D CV

Аннотация лекции и семинара: This lecture and seminar will discuss how diffusion models can be used for 3D computer vision tasks such as generation and reconstruction, and how they relate to methods like NeRF and Gaussian splatting. We will also look at how these models can be applied to problems like novel view synthesis, relighting, and camera pose or depth estimation.

Лектор и семинарист: [https://www.hse.ru/org/persons/885876805/ Mishan Aliev]

Локация: лекция - R205, семинар - R208.

Материалы: tba.

== Домашние задания ==

{| class="wikitable"
! № !! Домашнее задание !! Ссылка !! Дедлайн (жёсткий)
|-
| 1 || Tensor and DL Libraries || https://github.com/thecrazymage/DL2_HSE/tree/main/homeworks/homework_01 || 30 сентября, 23:59
|-
| 2 || Transformers for Named Entity Recognition || https://github.com/thecrazymage/DL2_HSE/tree/main/homeworks/homework_02 || 14 октября, 23:59
|-
| 3 || Image Segmentation || https://github.com/thecrazymage/DL2_HSE/tree/main/homeworks/homework_03 || 2 ноября, 23:59
|-
| 4 || Diffusion Models || https://github.com/thecrazymage/DL2_HSE/tree/main/homeworks/homework_04 || 18 ноября, 23:59

|}

== Соревнование ==

TBA

Глубинное обучение 2 2025

2025-11-13T17:29:59Z

Anapluslap: add materials for class 9

== Общая информация ==
Курс предназначен для студентов 4 курса ФКН ПМИ (МОП и КНАД).

Занятия проходят '''по понедельникам 14:40-17:40''' (переносы будут сообщаться в чате).

Полезные ссылки:
* Чат с обсуждением: https://t.me/+8dcjl4gHlyEwYzcy
* Репозиторий курса: https://github.com/thecrazymage/DL2_HSE
* Таблица с оценками: https://docs.google.com/spreadsheets/d/18rfZbf7Zsm-xmbiQn_eJVDXXbDfesbHT/edit?usp=sharing&ouid=115132401804687564737&rtpof=true&sd=true
* Anytask: https://anytask.org/course/1219

Формула итоговой оценки (округление арифметическое):
# МОП: Оитог = 0.25 * Осоревнование + 0.75 * ОДЗ,
# КНАД: Оитог = ОДЗ,
где ОДЗ - средняя оценка за практические домашние задания.

== Преподаватели и ассистенты ==

Кому писать, если кажется, что все пропало: [https://t.me/MishanAliev Мишан Алиев]

{| class="wikitable"
|-
! Группа !! Семинарист !! Ассистенты !! Чаты групп !! Инвайт в anytask
|-
| 221 (МОП) || [https://t.me/mightyneighbor Федя Великонивцев] || [https://t.me/vsem_paket Динар Саберов], [https://t.me/annvasileeva Анна Василева] || [https://t.me/+XzlFh0Y8GHs3NTMy МОП 221] || yvZZTIS
|-
| 222 (МОП) || [https://t.me/cocosinca Ева Неудачина] || [https://t.me/mat_os Александр Матосян], [https://t.me/kade1shvili Полина Кадейшвили] || [https://t.me/+PpxiIpF7wLYyMjE6 МОП 222] || W38CfZf
|-
| 223 (МОП) || [https://t.me/tutugarin Иван Ершов] || [https://t.me/Andrew_ut Андрей Уткин], [https://t.me/uzhedevyat Георгий Фатахов] || [https://t.me/+xRZkDO4TjiA1YjQ6 МОП 223] || ph91Jlz
|-
| 224 (МОП) || [https://t.me/Stepyndriy Степан Беляков] || [https://t.me/ponomarchuk_anna Анна Пономарчук], [https://t.me/bogsolntca Татьяна Яковлева] || [https://t.me/+XtcOG352A-IxMmY6 МОП 224] || iPwd342
|-
| КНАД || [https://t.me/burakevi4 Даня Бураков] || [https://t.me/anapluslap Анастасия Лапшина], [https://t.me/irbix7 Иван Галий] || [https://t.me/+JNPUgVxh6ophOGQy КНАД] || zTn4sRP
|}

== Лекции и семинары ==

'''Лекция / Семинар 1 (08.09).'''

Тема: Essentials of GPU, Deep Learning Bottlenecks, and Benchmarking Basics

Аннотация: In this session, we will explore the reasons behind the dominance of GPUs in Deep Learning and examine the common sources of performance bottlenecks in DL code. You will learn how to identify these bottlenecks using profiling tools and apply techniques to optimize and accelerate your code.

Лектор и семинарист: [https://www.hse.ru/org/persons/816100677/ Fedor Velikonivtsev]

Локация: лекция - R208, семинар - D102.

Материалы: [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/lecture_01.mp4 запись лекции], [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/seminar_01.mp4 запись семинара], [https://github.com/thecrazymage/DL2_HSE/tree/main/week_01 материалы].

'''Лекция / Семинар 2 (15.09).'''

Тема: On Transformers and Bitter Lesson

Аннотация: In this talk we’ll dive into the landscape of model architectures in deep learning, with a focus on the world around transformers. We’ll briefly recall what a transformer is, trace the evolution from encoder–decoder to encoder-only and decoder-only models, and touch on the rise of “efficient mixers” such as state space models, linear attention, and beyond. We’ll conclude by reflecting on the role of data and compute. The lecture is inspired in part by [http://incompleteideas.net/IncIdeas/BitterLesson.html The Bitter Lesson] and blends a brain dump with some entertaining insights from recent years of architectural exploration.

Лектор и семинарист: [https://www.hse.ru/org/persons/190912012/ Ivan Rubachev]

Локация: лекция - R206, семинар - R208.

Материалы: [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/lecture%2Bseminar_02.mp4 запись лекции и семинара], [https://github.com/thecrazymage/DL2_HSE/tree/main/week_02 материалы].

'''Лекция / Семинар 3 (22.09).'''

Тема: Modern LLMs essentials

Аннотация лекции: This week, we will discuss LLMs. We will discuss why they are so effective for text generation, how they can be applied to different NLP problems, and the risks they pose. You will learn the details of RLHF, PEFT, and RAG, which make LLMs robust in various cases.

Лектор: [https://www.hse.ru/org/persons/208533329/ Alexander Shabalin]

Локация: онлайн.

Аннотация семинара: In this seminar, we will explore the concept of LLM-based agents and how they extend the capabilities of modern language models. We will discuss function calling as a way to integrate external tools, chain-of-thought reasoning for structured problem solving, and reinforcement learning techniques for training agents.

Семинарист: [http://www.instagram.com/tugarin_vanya Ivan Ershov]

Локация: R208.

Материалы: [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/lecture_03.mp4 запись лекции], [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/seminar_03.mp4 запись семинара], [https://github.com/thecrazymage/DL2_HSE/tree/main/week_03 материалы].

'''Лекция / Семинар 4 (29.09).'''

Тема: Basics of Efficient LLM Training Infrastructure

Аннотация: In this lecture, we will study the basic rules that underpin the infrastructure for efficient large language model training. We will also examine common problems that may arise in this process and explore practical ways to address them.

Лектор и семинарист: [https://habr.com/ru/users/MichaelEk/ Michael Khrushchev]

Локация: лекция - G002, семинар - R208.

Материалы: [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/lecture%2Bseminar_04.mp4 запись лекции и семинара], [https://github.com/thecrazymage/DL2_HSE/tree/main/week_04 материалы].

'''Лекция / Семинар 5 (06.10).'''

Тема: Segmentation and Detection

Аннотация лекции: This week, we'll explore the evolution of object detection and segmentation — from R-CNN to real-time methods like YOLO-World and CLIP-based approaches. We'll examine how U-Net architectures have transcended computer vision to power neural operators and how to use diffusion models for segmentation tasks.

Аннотация семинара: We will implement architectures and train a semantic segmentation model. We will discuss regularization methods for convolutional layers. In addition, we will learn how to use pre-trained models and apply them.

Лектор и семинарист: [https://www.hse.ru/org/persons/956617478/ Alexander Oganov]

Локация: лекция - G002, семинар - R208.

Материалы: [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/lecture_05.mp4 запись лекции], [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/seminar_05.mp4 запись семинара], [https://github.com/thecrazymage/DL2_HSE/tree/main/week_05 материалы].

'''Лекция / Семинар 6 (13.10).'''

Тема: Segmentation and Detection 2

Аннотация лекции и семинара: In this lecture, we will continue in more detail about segmentation and detection.

Лектор: [https://szagoruyko.github.io/ Sergey Zagoruyko]

Семинарист: [https://www.hse.ru/org/persons/401628708/ Eva Neudachina]

Локация: лекция - G002, семинар - R208.

Материалы: [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/lecture_06.mp4 запись лекции], [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/seminar_06.mp4 запись семинара], [https://github.com/thecrazymage/DL2_HSE/tree/main/week_06 материалы].

'''Лекция / Семинар 7 (20.10).'''

Тема: Diffusion models 1

Аннотация лекции и семинара: Tomorrow's lecture and seminar will be devoted to an introduction to diffusion models. Diffusion models are currently the most popular approach to generative modeling due to their high-quality generation and diversity (mode coverage) of the learned distribution. The idea behind diffusion models is to consider the process of gradually transforming data into pure noise and construct its inverse in time, which will transform noise into data. In the lecture and seminar, we will work with noise processes and derive the classic [https://arxiv.org/abs/2006.11239 DDPM] model, which proposes to minimize the KL-divergence between the “true” reverse process that converts noise into data and the denoising process specified by the neural network. In the process, we will see that this procedure is equivalent to training a denoiser neural network that predicts a clean object from a noisy one. In addition, we will interpret the resulting denoising process: in it, each step corresponds to replacing part of the current noisy image with an (increasingly high-quality) prediction of the denoiser.

Лектор и семинарист: [https://www.hse.ru/org/persons/190910999/ Denis Rakitin]

Локация: лекция - G002, семинар - R208.

Материалы: [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/lecture_07.mp4 запись лекции], [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/seminar_07.mp4 запись семинара], [https://github.com/thecrazymage/DL2_HSE/tree/main/week_07 материалы].

'''Лекция / Семинар 8 (03.11).'''

Тема: Diffusion models 2

Аннотация лекции и семинара: In this lecture and seminar, we will continue our exploration of diffusion models. We will introduce the score function and score identity, present classifier and classifier-free guidance, and derive DDIM model.

Лектор и семинарист: [https://www.hse.ru/org/persons/190910999/ Denis Rakitin]

Локация: лекция - R205, семинар - R208.

Материалы: [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/lecture_08.mp4 запись лекции], [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/seminar_08.mp4 запись семинара], [https://github.com/thecrazymage/DL2_HSE/tree/main/week_08 материалы].

'''Лекция / Семинар 9 (11.11).'''

Тема: Diffusion models 3

Аннотация лекции: This lecture presents applied aspects of diffusion models, a class of generative methods that have demonstrated strong performance across images, text, and audio. We will review modern architectures, with a focus on [https://github.com/black-forest-labs/flux FLUX], and unpack the key principles of their design and training.

Лектор: [https://scholar.google.com/citations?user=o6pRm_gAAAAJ&hl=ru Nikita Starodubcev]

Аннотация семинара: The seminar will provide a hands-on, in-depth inspection of [https://github.com/PixArt-alpha/PixArt-alpha PixArt] model. Then ee will attempt further fine-tuning via [https://dreambooth.github.io/ DreamBooth], review common evaluation metrics, and compare different models and configurations.

Семинарист: [https://www.hse.ru/org/persons/401628708/ Eva Neudachina]

Локация: лекция - R205, семинар - R208.

Материалы: [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/lecture_09.mp4 запись лекции], [https://disk.yandex.ru/d/mzXlT0U3MzEZkQ/%D0%9F%D0%9C%D0%98/DL%202/seminar_09.mp4 запись семинара], [https://github.com/thecrazymage/DL2_HSE/tree/main/week_09 материалы].

== Домашние задания ==

{| class="wikitable"
! № !! Домашнее задание !! Ссылка !! Дедлайн (жёсткий)
|-
| 1 || Tensor and DL Libraries || https://github.com/thecrazymage/DL2_HSE/tree/main/homeworks/homework_01 || 30 сентября, 23:59
|-
| 2 || Transformers for Named Entity Recognition || https://github.com/thecrazymage/DL2_HSE/tree/main/homeworks/homework_02 || 14 октября, 23:59
|-
| 3 || Image Segmentation || https://github.com/thecrazymage/DL2_HSE/tree/main/homeworks/homework_03 || 2 ноября, 23:59
|-
| 4 || Diffusion Models || https://github.com/thecrazymage/DL2_HSE/tree/main/homeworks/homework_04 || 18 ноября, 23:59

|}

== Соревнование ==

TBA

Глубинное обучение 2 2025

2025-11-12T10:04:12Z

Anapluslap: add num of class

Глубинное обучение 2 2025

2025-11-12T09:53:30Z

Anapluslap: add dif models 3

Глубинное обучение 2 2025

2025-11-08T13:36:22Z

Anapluslap: add materials for class 8

Глубинное обучение 2 2025

2025-11-03T16:52:23Z

Anapluslap: add homework 4 Diffusion Models

Глубинное обучение 2 2025

2025-11-03T12:04:32Z

Anapluslap: add class 8

Глубинное обучение 2 2025

2025-10-27T17:00:21Z

Anapluslap: change deadline hw 3

Глубинное обучение 2 2025

2025-10-25T20:13:47Z

Anapluslap: add materials

Глубинное обучение 2 2025

2025-10-20T09:53:51Z

Anapluslap: add class 7 info

Глубинное обучение 2 2025

2025-10-19T15:09:11Z

Anapluslap: add class 5 materials

Глубинное обучение 2 2025

2025-10-14T16:42:48Z

Anapluslap: add homework 3

Глубинное обучение 2 2025

2025-10-13T10:16:54Z

Anapluslap:

Глубинное обучение 2 2025

2025-10-13T10:16:25Z

Anapluslap: add class 6 DL2

Глубинное обучение 2 2025

2025-10-05T13:19:35Z

Anapluslap: added new info on seminars/lectures 4,5

Глубинное обучение 2 2025

2025-09-29T09:45:49Z

Anapluslap: add class 4

Глубинное обучение 2 2025

2025-09-28T14:13:04Z

Anapluslap: change links

Глубинное обучение 2 2025

2025-09-28T14:12:00Z

Anapluslap: add materials 3rd week

Глубинное обучение 2 2025

2025-09-21T22:35:49Z

Anapluslap: /* Лекции и семинары */

Глубинное обучение 2 2025

2025-09-21T11:15:40Z

Anapluslap: add second class's materials DL2

Глубинное обучение 2 2025

2025-09-13T09:40:31Z

Anapluslap: add lecture seminar meterials for week 1

Глубинное обучение 2 2025

2025-09-13T09:28:57Z

Anapluslap: add lecture seminar 2

Глубинное обучение 2 2025

2025-09-07T20:58:15Z

Anapluslap: add DL2 description

== Общая информация ==
Курс предназначен для студентов 4 курса ФКН МОП.

Занятия проходят '''по понедельникам 16:20-19:30''' (переносы будут сообщаться в чате).

'''Лекторы''': [https://t.me/mightyneighbor Федя Великонивцев], [https://t.me/MishanAliev Мишан Алиев], Иван Рубачев, Саша Шабалин, Миша Хрущев, Саша Оганов, Сергей Загоруйко, Денис Ракитин, Дима Баранчук, Кирилл Хрыльченко

'''Семинаристы''': [https://t.me/burakevi4 Даня Бураков] (КНАД), [https://t.me/mightyneighbor Федя Великонивцев] (МОП 221), [https://t.me/cocosina Ева Неудачина] (МОП 222), [https://t.me/tutugarin Иван Ершов] (МОП 223), [https://t.me/Stepyndriy Степан Беляков] (МОП 224)

'''Ассистенты''': [https://t.me/anapluslap Анастасия Лапшина], [https://t.me/irbix7 Иван Галий] (КНАД), [https://t.me/vsem_paket Динар Саберов], [https://t.me/annvasileeva Анна Василева] (МОП 221), [https://t.me/mat_os Александр Матосян], [https://t.me/kade1shvili Полина Кадейшвили] (МОП 222), [https://t.me/Andrew_ut Андрей Уткин], [https://t.me/uzhedevyat Георгий Фатахов] (МОП 223), [https://t.me/ponomarchuk_anna Анна Пономарчук], [https://t.me/bogsolntca Татьяна Яковлева] (МОП 224)

[https://t.me/+8dcjl4gHlyEwYzcy '''Общий чат курса''']

'''Чаты групп''': [https://t.me/+JNPUgVxh6ophOGQy КНАД], [https://t.me/+XzlFh0Y8GHs3NTMy МОП 221], [https://t.me/+PpxiIpF7wLYyMjE6 МОП 222], [https://t.me/+xRZkDO4TjiA1YjQ6 МОП 223], [https://t.me/+XtcOG352A-IxMmY6 МОП 224]

[https://github.com/thecrazymage/DL2_HSE '''Репозиторий курса с материалами''']

== Оценки ==

Формула итоговой оценки:

ПМИ:
Оитог = 0.25 * Осорева + 0.75 * Одз. Округление арифметическое.

Одз - средняя оценка за практические домашние задания.

КНАД: ∑ ДЗi

== Сорева ==

Студенту предоставляется ограниченное число кадров одной 3D-сцены, после чего требуется сгенерировать изображения этой сцены с альтернативных ракурсов. Полученные результаты сравниваются с эталонными изображениями по классическим метрикам качества.

== Домашние задания ==

[https://github.com/puhsu/dl-hse/tree/main/2024/hw01-tensorlibs '''Домашнее задание №1''']

[https://github.com/mmp-practicum-team/mmp_dl_spring_2025/blob/main/Tasks/task3_5/task_03_5.ipynb '''Домашнее задание №2''']

[https://github.com/mmp-practicum-team/mmp_dl_spring_2025/blob/main/Tasks/task2/task_02.ipynb '''Домашнее задание №3''']

[https://github.com/yandexdataschool/deep_vision_and_graphics/blob/fall24/homework04/homework-part1-diffusion.ipynb '''Домашнее задание №4''']

[https://disk.yandex.ru/i/RJsps26HeG8ysQ '''Домашнее задание №5''']

== Лекции и семинары ==

'''Лекция / Семинар 1.'''
'''Тема:''' Essentials of GPU, Deep Learning Bottlenecks, and Benchmarking Basics

'''Аннотация:''' In this session, we will explore the reasons behind the dominance of GPUs in Deep Learning and examine the common sources of performance bottlenecks in DL code. You will learn how to identify these bottlenecks using profiling tools and apply techniques to optimize and accelerate your code.

Лектор и семинарист: [https://www.hse.ru/org/persons/816100677/ Fedor Velikonivtsev]

Локация: лекция - R208, семинар - D102.

[https://us06web.zoom.us/j/84219135654?pwd=BuXCj4CzvczH9uUE7IcfkPUfJvXdqf.1 '''Ссылка на занятие''']

Wiki ФКН

2025-09-06T01:12:29Z

Anapluslap: add new course (DL-2)

__NOTOC__
= Учебные курсы факультета компьютерных наук =

{| class="wikitable"
|-
! colspan="2" | <div style="text-align: center;">'''Навигация'''</div>
|-
! [[#bachelors|Курсы бакалавриата ФКН]]
| <div style="text-align:center">[[#AMI|ПМИ]] · [[#SE|ПИ]] · [[#DSBA|ПАД]] · [[#compds|КНАД]] · [[#EDA|ЭАД]] · [[#DRIP|ДРИП]] · [[#electives|майноры и факультативы]]</div>
|-
! colspan="2" | [[#DataCulture|Курсы в рамках проекта Data Culture]] · [[#masters|Курсы магистратуры ФКН]] · [[#other|Курсы других факультетов]] · [[#archive|Архив]]
|}

== Курсы за 2025/26 учебный год ==
{| class="wikitable courses"
|-

! width="21%" | 1 курс !! width="21%" | 2 курс !! width="21%" | 3 курс !! width="21%" | 4 курс !! rowspan="2" | '''майноры и факультативы'''
|-
| colspan="4" style="text-align: center;" | '''ПМИ'''
|-
|

'''М+'''

[[Линейная_алгебра_и_геометрия_на_ПМИ_2025/2026_(пилотный_поток) | Линейная алгебра и геометрия (пилотный поток)]]

'''М'''

[[Линейная_алгебра_и_геометрия_на_ПМИ_2025/2026_(основной_поток) | Линейная алгебра и геометрия (основной поток)]]

[[Математический_анализ_1_2025/26_(основной_поток) | Математический анализ-1 (ПМИ основной поток)]]

[[DM1PMIbase-2025-26 | Дискретная математика (ПМИ основной поток)]]

'''П'''

[[Язык программирования Python 2025/26 (основной поток) ]]

 

||

'''М+'''

[[Математический_Анализ_2_на_ПМИ_2025/26_(пилотный_поток) | Математический анализ 2 (пилотный поток)]]

'''М'''

[[Теория_вероятностей_на_ПМИ_2025/2026_(основной_поток) | Теория вероятностей (основной поток)]]

 

'''П'''

[[Язык_программирования_Python_(углубленный_курс) | Язык программирования Python (углубленный курс)]]
 

||

'''ПМИ / МОП'''

[[Машинное_обучение_1_25/26 | Машинное обучение 1]]

[[Математическая_статистика_2_2025/2026 | Математическая статистика 2]]

'''ПМИ / РС'''

[https://docs.google.com/document/d/1sHfi9_m4hx4y0iIgQEueMCr4qokLxh2Bvyzw_MdE2iI/edit?usp=sharing Распределенные системы]

'''ПМИ / ТИ'''

[[ NIS-TCS-25-26 | НИС Теоретическая информатика ]]

[[ PE101-25-26 | Perfomance engineering 101 ]]

 

||

[http://wiki.cs.hse.ru/%D0%93%D0%B5%D0%BD%D0%B5%D1%80%D0%B0%D1%82%D0%B8%D0%B2%D0%BD%D1%8B%D0%B5_%D0%BC%D0%BE%D0%B4%D0%B5%D0%BB%D0%B8_%D0%BD%D0%B0_%D0%BE%D1%81%D0%BD%D0%BE%D0%B2%D0%B5_%D0%B4%D0%B8%D1%84%D1%84%D1%83%D0%B7%D0%B8%D0%B8_(25/26) Генеративные модели на основе диффузии 25/26]

'''ПМИ / РС'''

[https://docs.google.com/document/d/1Q2gx-6ugSi9intAW6ZSsLZNrEvl7S5ClyAhn8EjJnRM/edit?usp=sharing НИС Распределенные системы 2]

'''ПМИ / ТИ'''

[[ NIS-TCS-25-26 | НИС Теоретическая информатика ]]

[[ PE101-25-26 | Perfomance engineering 101 ]]

[[ OWF-25-26 | Односторонние функции и их применения ]]

 

| rowspan="11" |


[[Введение_в_программирование_25/26|Введение в программирование. Питон 25/26]]

[[Введение_в_базы_данных_25/26|Введение в базы данных 25/26]]

[[Основы_глубинного_обучения_25/26|Основы глубинного обучения 25/26]]

[[Deep Learning-2 | Deep Learning-2 25/26]]

|-
| colspan="4" style="text-align: center;" | '''ПИ'''
|-
|

[[Алгебра_ПИ_2025-2026|Алгебра 25/26]]

[[Дискретная_математика_25/26|Дискретная математика 25/26]]

[[Математический_анализ_25/26|Математический анализ 25/26]]

||
 

||
 

|-
| colspan="4" style="text-align: center;" | '''ПАД'''
|-
|
'''1st year DSBA 2025/2026'''

[[Calculus 1 DSBA 2025/2026 | Calculus 1 DSBA (modules 1-4)]]

[[C++ Programming Language DSBA 2025/2026 | C++ Programming Language DSBA (modules 1-3)]]

[[English DSBA 2025/2026 | English Language DSBA (modules 1-4)]]

[[LAaG DSBA 2025/2026 | Linear Algebra and Geometry DSBA (modules 1-4)]]

[[Discrete Mathematics 1 DSBA 2025/2026 | Discrete Mathematics 1 DSBA (modules 1-3)]]

[[Russian History DSBA 2025/2026 | Russian History DSBA (modules 1-4)]]

[[Economics DSBA 2025/2026 | Economics DSBA (modules 2-3)]]

[[Foundations of Russian Statehood DSBA 2025/2026 | Foundations of Russian Statehood DSBA (module 3)]]

[[Algebra DSBA 2025/2026 | Algebra DSBA (module 4)]]

[[Algorithms and Data Structures 1 DSBA 2025/2026 | Algorithms and Data Structures 1 DSBA (module 4)]]

[[Python for Data Science DSBA 2025/2026 | Python for Data Science DSBA (module 4)]]

||
'''2nd year DSBA 2025/2026'''

[[Algorithms and Data Structures DSBA 2025/2026 | Algorithms and Data Structures DSBA (modules 1-3)]]

[[Discrete Mathematics 2 DSBA 2025/2026 | Discrete Mathematics 2 DSBA (modules 1-2)]]

[[Calculus 2 DSBA 2025/2026 | Calculus 2 DSBA (modules 1-2)]]

[[Probability Theory DSBA 2025/2026 | Probability Theory DSBA (modules 1-2)]]

[[Introduction to Micro and Macroeconomics DSBA 2025/2026 | Introduction to Micro and Macroeconomics DSBA (modules 1-4)]]

[[Computer Architecture and Operating Systems DSBA 2025/2026 | Computer Architecture and Operating Systems DSBA (modules 3-4)]]

[[Mathematical Statistics DSBA 2025/2026 | Mathematical Statistics DSBA (modules 3-4)]]

[[Machine Learning 1 DSBA 2025/2026 modules 3-4 | Machine Learning 1 DSBA (modules 3-4)]]

[[Differential Equations DSBA 2025/2026 | Differential Equations DSBA (modules (3-4)]]

'''Minors'''

[[Business and Management in Global Context DSBA 2025/2026 | Business and Management in Global Context DSBA (modules 1-4)]]

[[Introduction to Finance DSBA 2025/2026 | Introduction to Finance DSBA (modules 1-4)]]

||
'''3rd year DSBA 2025/2026'''

[[Databases DSBA 2025/2026 | Databases DSBA (modules 1-2)]]

[[Optimization Methods DSBA 2025/2026 | Optimization Methods DSBA (modules 1-2)]]

[[Machine Learning 1 DSBA 2025/2026 | Machine Learning 1 DSBA (modules 1-2)]]

[[Stochastic processes and applications DSBA 2025/2026 | Stochastic processes and applications DSBA (modules 1-2)]]

[[Time Series Analysis DSBA 2025/2026 | Time Series Analysis DSBA (modules 3-4)]]

[[Deep Learning DSBA 2025/2026 | Deep Learning DSBA (modules 3-4)]]

'''Minors'''

[[Introduction to Entrepreneurship DSBA 2025/2026 | Introduction to Entrepreneurship DSBA business minor (modules 1-2)]]

[[Information Systems Management DSBA 2025/2026 | Information Systems Management DSBA business minor (modules 3-4)]]

[[Econometrics DSBA 2025/2026 | Elements of Econometrics DSBA finance minor (modules 1-4)]]

'''Specialization Data Science in Business'''

[[Research Seminar "Data Analysis in Business" DSBA 2025/2026 | Research Seminar "Data Analysis in Business“ DSBA (modules 1-4)]]

[[Data Analysis in Business DSBA 2025/2026 | Data Analysis in Business DSBA (modules 1-2)]]

[[Investment Management for Business DSBA 2025/2026 | Investment Management for Business DSBA (modules 3-4)]]

'''Specialization Data Science in Finance'''

[[Research Seminar "Data Science in Financial Economics" DSBA 2025/2026 | Research Seminar "Data Science in Financial Economics" DSBA (modules 1-4)]]

[[Financial Mathematics DSBA 2025/2026 | Financial Mathematics DSBA (modules 1-2)]]

[[Risk Management in Bank DSBA 2025/2026 | Risk Management in Bank DSBA (modules 3-4)]]

'''Specialization Data Analysis in Applied Research'''

[[Research Seminar "Data Science in Scientific Research" DSBA 2025/2026 | Research Seminar "Data Science in Scientific Research" DSBA (modules 1-4)]]

[[Methods of Mathematical Modeling DSBA 2025/2026 | Methods of Mathematical Modeling DSBA (modules 1-2)]]

[[Applied Statistics for Machine Learning DSBA 2025/2026 | Applied Statistics for Machine Learning DSBA (modules 3-4)]]

'''Electives (modules 3-4)'''

[[Operations Research and Game Theory DSBA 2025/2026 | Operations Research and Game Theory DSBA]]

[[Advanced Statistical Methods DSBA 2025/2026 | Advanced Statistical Methods DSBA]]

[[Recommender Systems DSBA 2025/2026 | Recommender Systems DSBA]]

||
'''4th year DSBA 2025/2026'''

[[Strategy DSBA 2025/2026 | Strategy DSBA (modules 1-3)]]

[[Statistical Methods for Market Research DSBA 2025/2026 | Statistical Methods for Market Research DSBA (modules 2-3)]]

'''Electives block 1 (modules 1-3)'''

[[Core Management Concepts DSBA 2025/2026 | Core Management Concepts DSBA]]

[[Asset Pricing and Financial Markets DSBA 2025/2026 | Asset Pricing and Financial Markets DSBA]]

'''Electives block 2 (modules 1-2)'''

[[Introduction to the Study of Language for Computer Scientists DSBA 2025/2026 | Introduction to the Study of Language for Computer Scientists DSBA]]

[[Quantitative Finance DSBA 2025/2026 | Quantitative Finance DSBA]]

[[nformation Security Risk Management DSBA 2025/2026 | nformation Security Risk Management DSBA]]

[[Generative Models in Machine Learning DSBA 2025/2026 | Generative Models in Machine Learning DSBA]]

[[Natural Language Processing DSBA 2025/2026 | Natural Language Processing DSBA]]

'''Electives block 3 (module 3)'''

[[Information Systems DSBA 2025/2026 | Information Systems DSBA]]

[[Cоmputer Vision DSBA 2025/2026 | Cоmputer Vision DSBA]]

'''Research Seminar Electives (modules 1-3)'''

[[Research Seminar "Data Analysis in the Natural Sciences" DSBA 2025/2026 | Research Seminar "Data Analysis in the Natural Sciences" DSBA]]

[[Research Seminar "Data Science in Financial Markets" DSBA 2025/2026 | Research Seminar "Data Science in Financial Markets" DSBA]]

[[Research Seminar "Data analysis in complex systems" DSBA 2025/2026 | Research Seminar "Data analysis in complex systems"DSBA]]

|-
| colspan="4" style="text-align: center;" | '''КНАД'''
|-
|
[[Линейная алгебра КНАД 25/26]]

||
[[Алгоритмы_и_структуры_данных_2_КНАД_25/26 | Алгоритмы и структуры данных-2]]

||
 

||
 

|-
| colspan="4" style="text-align: center;" | '''ЭАД'''
|-
|
[[Язык программирования Python 2025/26 (ЭАД) ]]

[[DM1EAD-2025-26 | Дискретная математика 2025/26 (ЭАД)]]

[[Математический_анализ_1_2025/26_(основной_поток)_ЭАД | Математический анализ-1 (ЭАД)]]
||
 

||
 

||
 

|-
| colspan="4" style="text-align: center;" | '''ДРИП'''
|-
|
[[Ddip2529 | Линейная алгебра и геометрия]]

[[DM_DRIP-2025-26 | Дискретная математика 2025/26 (ДРИП)]]

||
 

||
 

||
 

|}

== Курсы в рамках проекта [https://www.hse.ru/dataculture/ Data Culture] ==

{| class="wikitable"
|-
! Осенний семестр !! Весенний семестр
|-
|
|
|}

== Курсы магистратуры ФКН ==
{| class="wikitable"
|-
! Ссылка !! Дисциплина !! Год обучения
|-
| [[ NIS-TCS-25-26 | НИС Теоретическая информатика ]] || НИС ТИ || СКН, 1-2 год
|}

== Курсы других факультетов ==

{| class="wikitable"
|-
! Дисциплина !! Курс !! Период
|-
|}

== Факультативы ==

{| class="wikitable"
|-
! Дисциплина !! Период
|-
|}

= Архив до 2024/25 учебного года включительно =
[[Wiki ФКН/Архив]]