Talk on "Novel First Order Bayesian Optimization "
Speaker: Dr. Prabuchandran K. J
Title: Novel First Order Bayesian Optimization
Abstract:Zeroth Order Bayesian Optimization (ZOBO) methods optimize an unknown function based on its black-box evaluations at the query locations. Unlike most optimization procedures, ZOBO methods fail to utilize gradient information even when it is available. On the other hand, First Order Bayesian Optimization (FOBO) methods exploit the available gradient information to arrive at better solutions faster. However, utilizing the gradient information in the FOBO methods incur severe computational cost. Further, the inherent sequential nature of the FOBO methods and the problem of noisy gradient information exacerbates the computational overhead as well as the underlying parameter estimation problem, limiting their wide applicability. To alleviate the aforementioned difficulties of FOBO methods, in this talk we propose a simple statistical model to leverage the gradient information and present novel FOBO methods that search for global optima effectively. Furthermore, we explore application of the proposed FOBO methods in the context of policy gradient reinforcement learning.
Speaker Bio: Prabuchandran K.J. is an Amazon-IISc Postdoctoral scholar at IISc, Bangalore. He holds Ph.D. from Department of Computer Science and Automation, IISc in the area of Reinforcement Learning. Post his PhD, Prabuchandran worked as Research Scientist at IBM Research Labs, India for an year and half on change detection algorithms for multivariate compositional data. His research lies in the intersection of reinforcement learning, stochastic control & optimization, Machine Learning, Bayesian Optimization and stochastic approximation algorithms. His research interest also focuses on utilizing techniques from these fields in solving problems arising in applications like wireless sensor networks, traffic signal control and social networks.
Event Date: 20th March, 2019(Wednesday)
Event Time: 02:30 PM to 03:30 PM
Venue: Room 23
IIT Dharwad, Karnataka