Structured Evolution with Compact Architectures for Scalable Policy Optimization

TL;DR

This paper introduces a structured random orthogonal matrix-based gradient estimation method for blackbox optimization, enabling the learning of compact, efficient policies with theoretical guarantees and practical advantages in robotics and high-dimensional spaces.

Contribution

The paper proposes a novel structured gradient estimator that improves accuracy and scalability, facilitating the training of compact policies with theoretical support and practical efficiency.

Findings

Achieves better policy quality with fewer parameters.

Provides faster training and inference for compact policies.

Solves robotics tasks with significantly fewer parameters than previous methods.

Abstract

We present a new method of blackbox optimization via gradient approximation with the use of structured random orthogonal matrices, providing more accurate estimators than baselines and with provable theoretical guarantees. We show that this algorithm can be successfully applied to learn better quality compact policies than those using standard gradient estimation techniques. The compact policies we learn have several advantages over unstructured ones, including faster training algorithms and faster inference. These benefits are important when the policy is deployed on real hardware with limited resources. Further, compact policies provide more scalable architectures for derivative-free optimization (DFO) in high-dimensional spaces. We show that most robotics tasks from the OpenAI Gym can be solved using neural networks with less than 300 parameters, with almost linear time complexity of the inference phase, with up to 13x fewer parameters relative to the Evolution Strategies (ES) algorithm introduced by Salimans et al. (2017). We do not need heuristics such as fitness shaping to learn good quality policies, resulting in a simple and theoretically motivated training mechanism.