Course Description
This course provides contents on Deep Generative Model(DGM) and Multi-Armed Bandit(MAB)/Bayesian Optimization(BO) in the field of Aritificial Intelligence with statistical foundations. DGM is a model to generate a virtual data instance and an inference on latent data distributions and its success is one key driver of the recent AI field. DGM has expanded from GAN and VAE to Flow, Energy-Based, and Score Matching models. From the perspective of probabilistic modeling, DGM shares the common ground with Multi-Armed Bandit, Stochastic Process, and Bayesian Optimization. MAB/BO is estimating the uncertainty of the observed data, and optimizing the decision-making with the inferred reward distribution. MAB has been utlized in automatic financial transaction systems and recommendation systems, and BO is the probabilistic machine learning of optimal control on AutoML and Robotics.