State of the Art Control of Atari Games Using Shallow Reinforcement Learning

TL;DR

This paper analyzes the principles behind Deep Q-Networks' success in Atari games, proposing a simple, effective representation that rivals DQN and offers a benchmark for future research.

Contribution

It introduces a linear representation capturing key features of DQN, reducing the need for game-specific learning and providing a reproducible benchmark.

Findings

The linear representation achieves performance comparable to DQN.

It offers insights into DQN's strengths and weaknesses.

Provides a generic, practical feature set for the ALE.

Abstract

The recently introduced Deep Q-Networks (DQN) algorithm has gained attention as one of the first successful combinations of deep neural networks and reinforcement learning. Its promise was demonstrated in the Arcade Learning Environment (ALE), a challenging framework composed of dozens of Atari 2600 games used to evaluate general competency in AI. It achieved dramatically better results than earlier approaches, showing that its ability to learn good representations is quite robust and general. This paper attempts to understand the principles that underlie DQN's impressive performance and to better contextualize its success. We systematically evaluate the importance of key representational biases encoded by DQN's network by proposing simple linear representations that make use of these concepts. Incorporating these characteristics, we obtain a computationally practical feature set that achieves competitive performance to DQN in the ALE. Besides offering insight into the strengths and weaknesses of DQN, we provide a generic representation for the ALE, significantly reducing the burden of learning a representation for each game. Moreover, we also provide a simple, reproducible benchmark for the sake of comparison to future work in the ALE.