Recent posts

Accepted Papers

Deep Reinforcement Learning Works - Now What?

It's safe to assume that deep reinforcement learning does indeed work. This is backed by recent trends which have achieved tremendous feats. An important question is — now what? In this post I question certain trends in deep RL research and propose some insights and solutions.

It's safe to assume that deep reinforcement learning does indeed work. This is backed by recent trends which have achieved tremendous feats. An important question is — now what? In this post I question certain trends in deep RL research and propose some insights and solutions.

Inverse Reinforcement Learning in Contextual MDPs

Real-world sequential decision problems often share two important properties – the reward function is often unknown, yet expert demonstrations can be acquired, and there often exists a static parameter, also known as the context, which determines certain aspects of the problem. In this work we formalize the Contextual Inverse Reinforcement Learning framework, propose several algorithms and analyze them both theoretically and empirically.

Distributional Policy Optimization: An Alternative Approach for Continuous Control

We propose a method for learning distributional policies, policies which are not limited to parametric distribution functions (e.g., Gaussian and Delta). This approach overcomes sub-optimal local extremum in continuous control regimes.

Action Robust Reinforcement Learning and Applications in Continuous Control

Action Robust is a special case of robustness, in which the agent is robust to uncertainty in the performed action. We show (theoretically) that this form of robustness has efficient solutions and (empirically) results in policies which are robust to common uncertainties in robotic domains.

Reward Constrained Policy Optimization

Learning a policy which adheres to behavioral constraints is an important task. Our algorithm, RCPO, enables the satisfaction of not only discounted constraints but also average and probabilistic, in an efficient manner.

A Deep Hierarchical Approach to Lifelong Learning in Minecraft

We propose a lifelong learning system that has the ability to reuse and transfer knowledge from one task to another while efficiently retaining the previously learned knowledge-base. Knowledge is transferred by learning reusable skills to solve tasks in Minecraft, a popular video game which is an unsolved and high-dimensional lifelong learning problem.