← Back to papers
2023 ICML ICML 2023

On Many-Actions Policy Gradient

Abstract

We study the variance of stochastic policy gradients (SPGs) with many action samples per state. We derive a many-actions optimality condition, which determines when many-actions SPG yields lower variance as compared to a single-action agent with proportionally extended trajectory. We propose Model-Based Many-Actions (MBMA), an approach leveraging dynamics models for many-actions sampling in the context of SPG. MBMA addresses issues associated with existing implementations of many-actions SPG and yields lower bias and comparable variance to SPG estimated from states in model-simulated rollouts. We find that MBMA bias and variance structure matches that predicted by theory. As a result, MBMA achieves improved sample efficiency and higher returns on a range of continuous action environments as compared to model-free, many-actions, and model-based on-policy SPG baselines.

🌉 Interdisciplinary Bridge — Machine Learning and Reinforcement Learning
🐝 Cross-Pollinator — Artificial Intelligence, Computer Science, Computer Vision, Data Science & Analytics, Deep Learning, Healthcare & Medicine, Interdisciplinary, Knowledge & Reasoning, Machine Learning, Mathematics & Optimization, Natural Language Processing, Reinforcement Learning, Robotics, Security & Privacy, Speech & Audio

Authors

Michal Nauman , Marek Cygan

Topics

Reinforcement Learning > Methods > Deep RL Reinforcement Learning > Methods > Policy Learning Machine Learning > Learning Types > Reinforcement Learning

Keywords

sample efficiency policy gradient model-based reinforcement learning variance reduction continuous action
Download PDF

Related papers

Orthogonality-Enforced Latent Space in Autoencoders: An Approach to Learning Disentangled Representations 2023
On the Convergence Rate of Gaussianization with Random Rotations 2023
Synthetic Prompting: Generating Chain-of-Thought Demonstrations for Large Language Models 2023
Generative Adversarial Symmetry Discovery 2023
Function-Space Regularization in Neural Networks: A Probabilistic Perspective 2023