On the Role of Iterative Computation in Reinforcement Learning

TL;DR

This paper formalizes compute-bounded policies in reinforcement learning, demonstrating that increased compute allows solving more complex tasks and better generalization, supported by theoretical proofs and extensive experiments.

Contribution

It introduces a minimal architecture for variable compute in RL and provides theoretical and empirical evidence of its advantages over traditional fixed-architecture policies.

Findings

More compute improves policy performance.

Increased compute enhances generalization to longer horizons.

Architecture outperforms standard networks with fewer parameters.

Abstract

How does the amount of compute available to a reinforcement learning (RL) policy affect its learning? Can policies using a fixed amount of parameters, still benefit from additional compute? The standard RL framework does not provide a language to answer these questions formally. Empirically, deep RL policies are often parameterized as neural networks with static architectures, conflating the amount of compute and the number of parameters. In this paper, we formalize compute bounded policies and prove that policies which use more compute can solve problems and generalize to longer-horizon tasks that are outside the scope of policies with less compute. Building on prior work in algorithmic learning and model-free planning, we propose a minimal architecture that can use a variable amount of compute. Our experiments complement our theory. On a set 31 different tasks spanning online and offline RL, we show that $(1)$ this architecture achieves stronger performance simply by using more compute, and $(2)$ stronger generalization on longer-horizon test tasks compared to standard feedforward networks or deep residual network using up to 5 times more parameters.