← Search

Ashwin Reddy

2 accepted papers

2021

Learning to Reach Goals via Iterated Supervised Learning

ICLR 2021oral

Current reinforcement learning (RL) algorithms can be brittle and difficult to use, especially when learning goal-reaching behaviors from sparse rewards. Although supervised imitation learning provides a simple and stable alternative, it requires access to demonstrations from a human supervisor. In…

2021

MURAL: Meta-Learning Uncertainty-Aware Rewards for Outcome-Driven Reinforcement Learning

ICML 2021spotlight

Exploration in reinforcement learning is, in general, a challenging problem. A common technique to make learning easier is providing demonstrations from a human supervisor, but such demonstrations can be expensive and time-consuming to acquire. In this work, we study a more tractable class of reinfo…

Cited by 46SourcePDFScholar