Current Projects
Project | 01 (in progress)
Model Selection for Offline Reinforcement Learning
Model selection in sequential decision making is generally considered very challenging. There has been relatively limited work in the generic setting until recently for some special cases. In this work we look for model selection methods in offline RL, a field that is commonly based on learned or prior models.
Project | 02 (in progress)
Learning Metrics Spaces for RL
A key component of any reinforcement learning algorithm is the underlying representation used by the agent. It has been shown that a metric on the state-action space can help construct efficient RL algorithms in non-tabular settings. In this work we attempt to learn such a metric using generative modeling approaches.
Project | 03 (in progress)
Natural Language State Representation for Reinforcement Learning
A natural way to describe what we observe, is through natural language. We implement a natural language state representation to learn and complete tasks. Experiments suggest that natural language based agents are more robust, converge faster and perform better than vision based agents, showing the benefit of using natural language representations for Reinforcement Learning.