The lecture presents theory behind machine learning. We will cover the following topics:
Introduction to PAC:
1. Introduction to PAC. Learnability of particular classes of concepts (DNF formulas, automata)
2. Occam's razor in the PAC context. PAC-reducibility.
Which concepts are lernable:
3. Growth functions and the VC-dimension.
4. Rademacher's complexity and the margin theory.
Leraning methods:
5. Support Vector Machines. Kernels.
6. Boosting.
Other learning frameworks:
7. Online learning.
8. Reinforcement learning.
9. Combination methods.
The lecture can be given in English or Polish; we will decide after the first lecture.
The lecture and classes will be held remotely in case of lockdown in the autumn 2021.
