Implicit Two-Tower Policies

TL;DR

Implicit Two-Tower policies introduce a structured approach to reinforcement learning that improves computational efficiency and performance by disentangling action and state processing, suitable for various action spaces.

Contribution

The paper proposes a novel Implicit Two-Tower architecture for RL policies that enhances efficiency and performance, especially in blackbox optimization contexts.

Findings

Outperforms unstructured implicit policies in 15 environments

Achieves substantial computational gains

Compatible with discrete and continuous actions

Abstract

We present a new class of structured reinforcement learning policy-architectures, Implicit Two-Tower (ITT) policies, where the actions are chosen based on the attention scores of their learnable latent representations with those of the input states. By explicitly disentangling action from state processing in the policy stack, we achieve two main goals: substantial computational gains and better performance. Our architectures are compatible with both: discrete and continuous action spaces. By conducting tests on 15 environments from OpenAI Gym and DeepMind Control Suite, we show that ITT-architectures are particularly suited for blackbox/evolutionary optimization and the corresponding policy training algorithms outperform their vanilla unstructured implicit counterparts as well as commonly used explicit policies. We complement our analysis by showing how techniques such as hashing and lazy tower updates, critically relying on the two-tower structure of ITTs, can be applied to obtain additional computational improvements.