Learning to Reweight Imaginary Transitions for Model-Based Reinforcement Learning

TL;DR

This paper introduces a meta-gradient reweighting method for imaginary transitions in model-based RL, improving training stability and performance when the learned dynamics model is imperfect.

Contribution

It proposes a novel adaptive reweighting scheme for imaginary trajectories using a meta-gradient approach, enhancing model-based RL robustness.

Findings

Outperforms state-of-the-art RL algorithms on multiple tasks.

Effectively reduces negative impact of inaccurate imaginary trajectories.

Visualization confirms the importance of reweighting for training stability.

Abstract

Model-based reinforcement learning (RL) is more sample efficient than model-free RL by using imaginary trajectories generated by the learned dynamics model. When the model is inaccurate or biased, imaginary trajectories may be deleterious for training the action-value and policy functions. To alleviate such problem, this paper proposes to adaptively reweight the imaginary transitions, so as to reduce the negative effects of poorly generated trajectories. More specifically, we evaluate the effect of an imaginary transition by calculating the change of the loss computed on the real samples when we use the transition to train the action-value and policy functions. Based on this evaluation criterion, we construct the idea of reweighting each imaginary transition by a well-designed meta-gradient algorithm. Extensive experimental results demonstrate that our method outperforms state-of-the-art model-based and model-free RL algorithms on multiple tasks. Visualization of our changing weights further validates the necessity of utilizing reweight scheme.