"Interactive Learning in Structured and Partially Observable Domains" 
Kamyar Azizzadenesheli (Caltech)
Monday, March 11, 2019, 2:00pm
EBU3B, Room 4258
Abstract:
Interactive Learning (IL) is a prominent machine learning paradigm that explores how intelligent autonomous systems learn to make improved sequential decisions through multiple rounds of interactions with the real world. IL-based systems have a wide range of applications including robotics, health-care, and marketing. However, for these systems to meet the practical requirements of the related application domains, there is a perennial need for developing efficient algorithms. In this talk, I will elaborate on how my work finds principled ways of designing such algorithms in the contexts of Reinforcement Learning and Domain Adaptation. I will present novel exploration/exploitation reinforcement learning algorithms for Partially Observable Markov Decision Processes (POMDP). Next, I will discuss a new algorithm for domain adaptation in the presence of shifts in the label distribution, which has applications in disease diagnosing and cloud service providing. I will conclude the talk with applications of IL in drone navigation, multi-agent swarm, recommendation systems, resource allocation, distributed optimization, and video games.