This class is intended for advanced machine learning students who are interested in doing research on learning theory. Pre-requisites for this class are basic machine learning, an advanced algorithms class, and some prior exposure to learning theory.
In the first few lectures, we will cover some standard models and concepts in learning theory. The next few lectures will be on some special topics at the forefront of active research in learning theory. Specifically, we will look at the following topics:
- Stochastic Gradient Methods
- Active Learning
- Spectral Learning
- Domain Adaptation
- Privacy Preserving Learning
There is no textbook for this class. Some suggested references for the first few lectures are:
- M. Kearns and U. Vazirani, Introduction to Computational Learning Theory, MIT Press, 1997.
- L. Devroye, A. Gyorfi, G. Lugosi, A probabilistic theory of pattern recognition , Springer.