CANCELLED • Seminar • Scientific Computation — Learning Under Uncertainty: Jumping Out of the Traditional Stochastic Optimization Framework

Please note: This seminar has been cancelled

Niao He, Department of Industrial Systems Engineering and Coordinated Science Laboratory
University of Illinois at Urbana-Champaign

Uncertainty penetrates in every corner of machine learning, ranging from data and adversary uncertainty, model uncertainty, all the way to dynamics uncertainty, even task uncertainty, and beyond. When faced with complicated machine learning tasks under various forms of uncertainty, the traditional empirical risk minimization framework, along with the rich off-the-shelf stochastic optimization algorithms, may no longer be applicable. This calls for new frameworks, algorithms, and principles for handling the uncertainty and making learning effective.

In this talk, I will introduce a unified optimization framework, called conditional stochastic optimization, that covers a wide spectrum of learning paradigms under uncertainty, including reinforcement learning, meta-learning, adversarial machine learning. 

I will discuss the algorithms and sample complexities for solving this problem under various structural assumptions on smoothness and (non)convexity, both for the rich and limited sampling schemes. I will discuss how these optimization frameworks and perspectives can be leveraged to build theoretically-sound and practically-efficient algorithms for reinforcement learning, meta-learning, etc.

In addition, I will discuss some other optimization and control perspectives for understanding the convergence of classical and modern reinforcement learning algorithms (such as the standard Q-learning and deep Q-learning). 

Bio: Niao He is an assistant professor in the Department of Industrial Systems Engineering and Coordinated Science Laboratory at the University of Illinois at Urbana-Champaign (UIUC). She is also a core member of the Illinois Institute of Data Science and Dynamical Systems. Before joining Illinois, she received her Ph.D. degree in Operations Research from Georgia Institute of Technology in 2015 and B.S. degree in Mathematics from University of Science and Technology of China in 2010. 

Her research interests are at the interface of large-scale optimization and machine learning, with a primary focus on reinforcement learning, probabilistic inference, and learning under uncertainty. She is a recipient of the Best Paper Award at AISTATS 2016, the NSF CISE CRII award, the NCSA Faculty Fellow, and the CAS Fellow. She has also been on the List of Teachers Ranked as Excellent at UIUC multiple times.