English

Bootstrapped Meta-Learning

Machine Learning 2022-03-17 v2 Artificial Intelligence Machine Learning

Abstract

Meta-learning empowers artificial intelligence to increase its efficiency by learning how to learn. Unlocking this potential involves overcoming a challenging meta-optimisation problem. We propose an algorithm that tackles this problem by letting the meta-learner teach itself. The algorithm first bootstraps a target from the meta-learner, then optimises the meta-learner by minimising the distance to that target under a chosen (pseudo-)metric. Focusing on meta-learning with gradients, we establish conditions that guarantee performance improvements and show that the metric can control meta-optimisation. Meanwhile, the bootstrapping mechanism can extend the effective meta-learning horizon without requiring backpropagation through all updates. We achieve a new state-of-the art for model-free agents on the Atari ALE benchmark and demonstrate that it yields both performance and efficiency gains in multi-task meta-learning. Finally, we explore how bootstrapping opens up new possibilities and find that it can meta-learn efficient exploration in an epsilon-greedy Q-learning agent, without backpropagating through the update rule.

Keywords

Cite

@article{arxiv.2109.04504,
  title  = {Bootstrapped Meta-Learning},
  author = {Sebastian Flennerhag and Yannick Schroecker and Tom Zahavy and Hado van Hasselt and David Silver and Satinder Singh},
  journal= {arXiv preprint arXiv:2109.04504},
  year   = {2022}
}

Comments

Published at ICLR 2022. 37 pages, 19 figures, 9 tables