The goal is to establish mathematical theory for deep learning, in order
to explain its practical success.
The goal is to develop machine learning models for causal inference and methods that are robust under various scenarios such as noises, adversarial attack and out of distribution data.
The goal is to develop convex and nonconvex optimization algorithms, large scaled and distributed training algorithms, automatic tuning of machine learning models and efficient deployment.
The goal is to develop tools and apply learning algorithms
to various applications such as natural language processing,
computer vision, and games, etc.