Jalaj Bhandari
 |
PhD student, |
About me
I am a PhD student in Operations Research at Columbia University, working with Prof. Garud Iyengar and Prof. Daniel Russo. My thesis work explores foundations of modern Reinforcement Learning (RL) algorithms using ideas from optimization theory. In the past, I have done research work with Prof. John Cunningham on Bayesian Machine learning, specifically in designing computationally efficient Markov Chain Monte Carlo (MCMC) algorithms for posterior sampling. I am also excited in applying RL and machine learning methods to problems of practical interest, for example in the areas of healthcare, neuroscience, autonomous systems, personalized Ads and more.
Prior to Columbia, I graduated from India Institute of Technology (IIT), Delhi in 2012 with a B.Tech in Industrial Engineering and Operations Research. I briefly worked in the finance industry from 2012-2013.
Recent Publications
Global Optimality Guarantees For Policy Gradient Methods [pdf]
Jalaj Bhandari and Daniel Russo
OptRL workshop, NeurIPS 2019; RLDM 2019
Journal version in preparation
A Finite Time Analysis of Temporal Difference Learning With Linear Function Approximation [pdf]
Jalaj Bhandari, Daniel Russo and Raghav Singal
Conference on Learning Theory (COLT), 2018; OptRL workshop, NeurIPS 2019; RLDM 2019
Journal version accepted to Operations Research with minor revisions.
Annular Augmentation Sampling [pdf]
Francois Fagan, Jalaj Bhandari and John P. Cunningham
AISTATS, 2017 as full oral presentation (top 5%).
Elliptical Slice Sampling with Expectation Propagation [pdf]
Francois Fagan, Jalaj Bhandari and John P. Cunningham
UAI, 2016.