Pre-requisites: IE 502 or IE 621 or IE 615 or equivalent or Instructor's approval
Course Contents:
Classification loss functions and their risks; cost sensitive classification; Tong Zhang inequality; cost sensitive classification calibration; Generative Adversarial Networks (GANs) and some of their applications; mode collapse, training instability and other issues; Optimal transport, Wasserstein and related metrics; Wasserstein GANs and other variants; Learning representations; Review of Markov decision models; reinforcement learning; Q-learning, actor-critic algorithms, function approximations and related algorithms; deep RL; some applications; Reinforcement learning in random environment; change point detection algorithms; divergence based Markov decision models and related topics.
References:
- Ingo Steinwart and Andreas Chirstmann, Support Vector Machines, Springer Science+Business Media, 2008
- Richard Sutton and Andrew G. Barton, Reinforcement Learning, An Introduction, Second Edition, MIT Press, 2018
- Dimitri P. Bertsekas, Reinforcement Learning and Optimal Control, MIT Press, 2019
- Dimitri P. Bertsekas and John Tistsiklis, Neuro-Dynamic Programming, Athena Scientific, 1996
- Recent Open Literature