# Object Exchangeability in Reinforcement Learning: Extended Abstract

## TL;DR

This paper introduces an attention-based input representation method for deep reinforcement learning that is invariant to input ordering, significantly improving sample efficiency and enabling solutions to previously intractable problems.

## Contribution

The paper proposes a novel attention-based representation technique that reduces the search space and enhances sample efficiency in reinforcement learning tasks.

## Key findings

- Improved sample efficiency in policy gradient methods.
- Reduced search space by a factor of m! for m objects.
- Ability to solve complex problems previously intractable.

## Abstract

Although deep reinforcement learning has advanced significantly over the past several years, sample efficiency remains a major challenge. Careful choice of input representations can help improve efficiency depending on the structure present in the problem. In this work, we present an attention-based method to project inputs into an efficient representation space that is invariant under changes to input ordering. We show that our proposed representation results in a search space that is a factor of m! smaller for inputs of m objects. Our experiments demonstrate improvements in sample efficiency for policy gradient methods on a variety of tasks. We show that our representation allows us to solve problems that are otherwise intractable when using naive approaches.