MinAtar: An Atari-Inspired Testbed for Thorough and Reproducible Reinforcement Learning Experiments

Abstract

The Arcade Learning Environment (ALE) is a popular platform for evaluating reinforcement learning agents. Much of the appeal comes from the fact that Atari games demonstrate aspects of competency we expect from an intelligent agent and are not biased toward any particular solution approach. The challenge of the ALE includes (1) the representation learning problem of extracting pertinent information from raw pixels, and (2) the behavioural learning problem of leveraging complex, delayed associations between actions and rewards. Often, the research questions we are interested in pertain more to the latter, but the representation learning problem adds significant computational expense. We introduce MinAtar, short for miniature Atari, a new set of environments that capture the general mechanics of specific Atari games while simplifying the representational complexity to focus more on the behavioural challenges. MinAtar consists of analogues of five Atari games: Seaquest, Breakout, Asterix, Freeway and Space Invaders. Each MinAtar environment provides the agent with a 10 × 10 × n binary state representation. Each game plays out on a 10 × 10 grid with n channels corresponding to gamespecific objects, such as ball, paddle and brick in the game Breakout. To investigate the behavioural challenges posed by MinAtar, we evaluated a smaller version of the DQN architecture as well as online actor-critic with eligibility traces. With the representation learning problem simplified, we can perform experiments with significantly less computational expense. In our experiments, we use the saved compute time to perform step-size parameter sweeps and more runs than is typical for the ALE. Experiments like this improve reproducibility, and allow us to draw more confident conclusions. We hope that MinAtar can allow researchers to thoroughly investigate behavioural challenges similar to those inherent in the ALE. MotivationThe Arcade Learning Environment (Bellemare, Naddaf, Veness, & Bowling, 2013) (ALE) has become widely popular as a testbed for reinforcement learning (RL), and other artificial intelligence algorithms, due in part to the following features:• Nuanced exploration challenges: Atari games provide varying degrees of exploration challenge. In Seaquest, for example, an agent must learn to collect divers and go up for air to avoid running out of oxygen. Neither of these provide immediate reward, but failing to learn this will limit the return to what can be obtained in the time it takes for oxygen to run out. In Breakout, a strong strategy involves trapping the ball between the ceiling and the Preprint. Under review.

Two-Sample Testing in Reinforcement Learning

“…We select the MinAtar (Young and Tian, 2019) environments to test the proposed Deep RL algorithms. MinAtar is a testbed incorporating several Atari games from the Arcade Learning Environment (Bellemare et al, 2013) with a reduced state-representation.…”

“…The platform incorporates sticky actions (Machado et al, 2018) and is designed to enable thorough algorithmic comparisons due to reduced computation times. Following Young and Tian (2019), the network structure consists of a convolutional and a fully-connected layer. The remaining hyperparameters match Young and Tian (2019), except that we use the Adam (Kingma and Ba, 2014) optimizer, which led to much more stable results during our experiments.…”

“…Following Young and Tian (2019), the network structure consists of a convolutional and a fully-connected layer. The remaining hyperparameters match Young and Tian (2019), except that we use the Adam (Kingma and Ba, 2014) optimizer, which led to much more stable results during our experiments. Appendix B contains the full list of specifications.…”

“…Since the TE is a special case of the KE with κ(T ) = I(T ≥ z α ), the above formula is also applicable for TE. (Young and Tian, 2019). All algorithms were implemented using PyTorch (Paszke et al, 2019) and the computation was performed on Intel(R) Xeon(R) CPUs E5-2680 v3 (12 cores) @ 2.50GHz.…”

Neural Network Optimization for Reinforcement Learning Tasks Using Sparse Computations

“…Authors of [12] explored this phenomenon in detail for both DQN and PPO algorithm. However, they studied it on small MinAtar subtasks [13] (MinAtar is a simplified version of classic AtariGames).…”