Lecture 11 slides updated on May 18.
The lectures are given in English to all students. Lecture slides up to the latest lecture: smu-slides.pdf.
The book links below pointing to Springer Link let you download the PDFs if accessing from the CVUT domain. Don't get confused by the indexing in the Universal AI book: it is different from ours because Hutter assumes the agent-env interaction to start with an agent's action whereas we start with a percept from the environment (for us, this is more natural in the reinforcement and concept learning scenarios).
Lec # | Lecture date | Slides available | Online Meeting | (Planned) Contents | Additional Reading |
---|---|---|---|---|---|
1 | 17.2 | 17.2 | General framework, passive reinforcement learning, DUE and ADP agents | AIMA book Chapter 21, RL Book | |
2 | 24.2 | 24.2 | TD agent, active R/L, Q-learning | as above | |
2.3 | (school closed) | ||||
3 | 9.3 | 9.3 | SARSA agent, state representation, policy search, Bayesian approach | as above (except Bayesian) | |
4 | 23.3 | 30.3 16:15 | Finish Bayesian approach. Universal sequence prediction, AIXI agent; Intro to concept learning. | Universal AI Book Chapter 1 (optionally other chapters for deeper understanding); supplementary web | |
5 | 30.3 | 6.4 16:15 | Online concept learning, mistake-bound model | numerous sources here | |
6 | 6.4 | 13.4 16:15 | Batch-learning, PAC-learning model | as above | |
7 | 17.4 | 20.4. 16:15 | Finish PAC-learning. Learning relational concepts | ILP Book, Logical Learning Book | |
8 | 26.4 | 27.4 16:15 | Reduction, Learning with Background Knowledge | as above | |
9 | 4.5 | 4.5 16:15 | Functions, Inductive Logic Programming, Structured Output | as above + AIMA book Ch. 9.4.2 and 19.5; ILP wiki | |
10 | 11.5 | 11.5 16:15 | Bayesian networks - Conditional independence, Inference | AIMA book Ch. 14 | |
11 | 18.5 | 18.5 16:15 | Bayesian networks - MAP inference, parameter learning | AIMA book Ch. 20.2.4 (MAP Inference not in AIMA) | |
|
Consultations: please ask your questions in the forum and/or at the online meeting.