Online Meta-learning by Parallel Algorithm Competition

TL;DR

This paper introduces OMPAC, a parallel algorithm competition approach for online meta-learning in reinforcement learning, which adaptively tunes meta-parameters to improve performance in complex tasks.

Contribution

The paper presents a novel parallel meta-learning method, OMPAC, that dynamically adapts meta-parameters during reinforcement learning, outperforming state-of-the-art results in various games.

Findings

Improved results in stochastic SZ-Tetris and Tetris by 31% and 84%.

Enhanced deep Sarsa(λ) agents in Atari games by over 62%.

Demonstrated adaptive meta-parameter tuning during learning.

Abstract

The efficiency of reinforcement learning algorithms depends critically on a few meta-parameters that modulates the learning updates and the trade-off between exploration and exploitation. The adaptation of the meta-parameters is an open question in reinforcement learning, which arguably has become more of an issue recently with the success of deep reinforcement learning in high-dimensional state spaces. The long learning times in domains such as Atari 2600 video games makes it not feasible to perform comprehensive searches of appropriate meta-parameter values. We propose the Online Meta-learning by Parallel Algorithm Competition (OMPAC) method. In the OMPAC method, several instances of a reinforcement learning algorithm are run in parallel with small differences in the initial values of the meta-parameters. After a fixed number of episodes, the instances are selected based on their performance in the task at hand. Before continuing the learning, Gaussian noise is added to the meta-parameters with a predefined probability. We validate the OMPAC method by improving the state-of-the-art results in stochastic SZ-Tetris and in standard Tetris with a smaller, 10$\times$10, board, by 31% and 84%, respectively, and by improving the results for deep Sarsa($\lambda$) agents in three Atari 2600 games by 62% or more. The experiments also show the ability of the OMPAC method to adapt the meta-parameters according to the learning progress in different tasks.