Machine Learning Methods
| Code | Completion | Credits | Range | Language |
|---|---|---|---|---|
| BECM36MLM | Z,ZK | 6 | 2P+2C | English |
- Course guarantor:
- Filip Železný
- Lecturer:
- Ondřej Kuželka, Gustav Šír, Filip Železný
- Tutor:
- Martin Krutský, Ondřej Kuželka, Jakub Peleška, Gustav Šír
- Supervisor:
- Department of Computer Science
- Synopsis:
-
Students will get familiar with advanced machine learning methods (MLM) that go beyond common data domains (vision, text) taught in the other courses (e.g., BE4M33MPV, BECM36NLPT). They will learn techniques that work well for tabular and structured data (e.g., relational databases), including rule/tree ensembles, graph neural networks, and other advanced approaches aimed at complex learning problems.
Additionally, the course will also teach students methods for model interpretability, the basics of causality, and reinforcement learning.
- Requirements:
-
This is an advanced ML course that assumes at least some prior knowledge of ML (e.g., B4B33RPZ, BECM33MLF), data representation (e.g., B0B36DBS, B0B01LGR), and deep learning (e.g., BECM33DPL).
- Syllabus of lectures:
-
1. Learning from Tabular data
2. Learning from Structured data
3. Graph Neural Networks
4. Relational Deep Learning
5. Neural-Symbolic Learning
6. Learning with Large Language Models
7. Interpretability of ML Models
8. Potential outcomes - Rubin-Neyman causal model, uplift modeling
9. Intro to Pearls causality
10. A/B tests and multi-armed bandit problems, UCB algorithm.
11. Bayesian bandits (Thompson sampling). Contextual bandits.
12. Markov decision processes
13. Tabular RL: Q-Learning and SARSA
14. Deep RL: Deep Q-learning. Policy gradient.
- Syllabus of tutorials:
-
1. Learning from Tabular data
2. Learning from Structured data
3. Graph Neural Networks
4. Relational Deep Learning
5. Neural-Symbolic Learning
6. Learning with Large Language Models
7. Interpretability of ML Models
8. Potential outcomes - Rubin-Neyman causal model, uplift modeling
9. Intro to Pearls causality
10. A/B tests and multi-armed bandit problems, UCB algorithm.
11. Bayesian bandits (Thompson sampling). Contextual bandits.
12. Markov decision processes
13. Tabular RL: Q-Learning and SARSA
14. Deep RL: Deep Q-learning. Policy gradient.
- Study Objective:
- Study materials:
- Note:
- Further information:
- https://cw.fel.cvut.cz/b252/courses/becm36mlm/start
- Time-table for winter semester 2025/2026:
- Time-table is not available yet
- Time-table for summer semester 2025/2026:
-
06:00–08:0008:00–10:0010:00–12:0012:00–14:0014:00–16:0016:00–18:0018:00–20:0020:00–22:0022:00–24:00
Mon Tue Wed Thu Fri - The course is a part of the following study plans:
-
- prg.ai Master (compulsory course in the program)