Deep Learning 2 HSE

This repo contains lectures slides, seminars notebooks and homeworks for the "Deep Learning 2" course at the Faculty of Computer Science of HSE University.

Details about the course organization can be found at the wiki page (in Russian).

General information

1. The course provides 4 homework assignments and 2 practice tasks.
2. Evaluation formula (arithmetic rounding):
   • MOP and AI360: $$S = \text{round}\left(0.4\cdot\text{Practice} + 0.4\cdot\text{HW} + 0.2\cdot\text{Final Test}\right)$$
   • KNAD: $$S = \text{round}\left(0.7\cdot\text{HW} + 0.3\cdot\text{Final Test}\right)$$
3. Course components:
   • Homeworks: 4 homework assignments in total.
   • Practice (for AI360 & MOP):
     • Participation in a competition.
     • Task transferred from another course.
   • Final Test: questions covering one lecture block.

Syllabus

1.  DL 1 compressed
2.  Essentials of GPU, Deep Learning Bottlenecks, and Benchmarking Basics
3.  The Evolution of Transformers
4.  RLHF and LLM Agents
5.  Practical Challenges in Agentic Reinforcement Learning for LLMs
6.  Segmentation and Detection

Homeworks

1.  Homework 1
1.  Homework 2

Acknowledgments

I would like to express my gratitude to all our guest lecturers and seminarians for agreeing to participate and share their expertise. A special thanks to Maxim Kodryan — his mere existence was contribution enough.

License

The content of lectures and assignments is distributed under the Apache 2.0 license: you can use and redistribute it for any purposes, as long as you refer to this course as the origin of the content.