Hello world!
Here is an updated version of the webpage
My name is Reazul Hasan Russel. I am a PhD candidate at the department of Computer Science, University of New Hampshire . I am working under the supervision of Professor Dr. Marek Petrik. I am a member of Reinforcement Learning and Robustness Lab.
I completed my B.Sc. in Computer Science & Engineering from Southeast University Bangladesh in early 2012. I joined Samsung R&D Institute Bangladesh (SRBD) just after my graduation, worked there as a Software Engineer and then Sr. Software Engineer. I contributed on research and development of several flagship samsung products till mid 2015.
I then started my PhD in Computer Science and currently working toward that. I obtained my Masters as a byproduct in Spring 2018.
Research Interest
I am interetsed to apply and deploy Reinforcement Learning algorithms into real world domains. Applying RL to solve real world problems has several challenges, including without limitation:
- Training off-line from fixed logs generated by running a baseline policy.
- Sample amount is often limited and exploration is not allowed due to high cost of failure.
- Computed policy has to meet some safety requirements and needs to offer some kind of performance guarantees.
- State and action spaces are very high dimensional, continuous & often partially observable.
In my reserach, I care about developing reasonable solutions to these challenges. I am interested in model-based batch Reinforcement Learning, in particular computing robust and safe model based solutions for RL problems with fixed & limited dataset. From data, I not only learn models, I also learn confidence levels. These confidence levels reside in very high dimensional space and are usually known as ambiguity sets. These ambiguity sets then drive a robust optiization technique that maximizes a lower bound on the worst-case return. My research involves optimizing ambiguity sets to compute tight and realistic lower bound estimates. I also care about robust exploration techniques in case exploration is feasible in a domain.
Contact Information
Best way to reach me is email: rrussel at cs.unh.edu
Geographically, I'm normally distributed around my research lab, pointed by the marker below: