I am a Research Assistant at the University of Alberta working with Martha White in the RLAI Lab.

My current research centers on online reinforcement learning (RL), specifically developing methods that dynamically tune hyperparameters during allocated agent-environment interactions. I have also contributed to designing environments that are partially observable, never-ending, and continually changing, in order to evaluate the agent’s adaptability, similar to how natural intelligences operate. I also study how rich, diverse, task-agnostic observations can enhance an agent’s abilities across simple and complex RL environments.

I have recently developed a keen interest in quantifying the amount of information an agent needs for efficient decision-making, especially in a partially observable setting. I am interested in exploring capacity-constrained agents - agents with constraints on the information they can get about the world they live in.

Contact: hakhverd [at] ualberta [dot] ca