Evolution Strategies as a Scalable Alternative to Reinforcement Learning

03/10/2017
by   Tim Salimans, et al.
0

We explore the use of Evolution Strategies (ES), a class of black box optimization algorithms, as an alternative to popular MDP-based RL techniques such as Q-learning and Policy Gradients. Experiments on MuJoCo and Atari show that ES is a viable solution strategy that scales extremely well with the number of CPUs available: By using a novel communication strategy based on common random numbers, our ES implementation only needs to communicate scalars, making it possible to scale to over a thousand parallel workers. This allows us to solve 3D humanoid walking in 10 minutes and obtain competitive results on most Atari games after one hour of training. In addition, we highlight several advantages of ES as a black box optimization technique: it is invariant to action frequency and delayed rewards, tolerant of extremely long horizons, and does not need temporal discounting or value function approximation.

READ FULL TEXT

page 1

page 2

page 3

page 4

05/14/2014

COCOpf: An Algorithm Portfolio Framework

Algorithm portfolios represent a strategy of composing multiple heuristi...
12/08/2020

GPU Accelerated Exhaustive Search for Optimal Ensemble of Black-Box Optimization Algorithms

Black-box optimization is essential for tuning complex machine learning ...
05/24/2022

Regret-Aware Black-Box Optimization with Natural Gradients, Trust-Regions and Entropy Control

Most successful stochastic black-box optimizers, such as CMA-ES, use ran...
06/21/2019

Evolution Attack On Neural Networks

Many studies have been done to prove the vulnerability of neural network...
04/09/2022

Distributed Evolution Strategies for Black-box Stochastic Optimization

This work concerns the evolutionary approaches to distributed stochastic...
10/09/2020

Instance Weighted Incremental Evolution Strategies for Reinforcement Learning in Dynamic Environments

Evolution strategies (ES), as a family of black-box optimization algorit...
10/11/2018

Multi-Strategy Coevolving Aging Particle Optimization

We propose Multi-Strategy Coevolving Aging Particles (MS-CAP), a novel p...

Code Repositories

evolution-strategies-exploration

Contains implementation of: Tim Salimans, Jonathan Ho, Xi Chen, and Ilya Sutskever. “Evolution Strategies as a Scalable Alternative to Reinforcement Learning”. Arxiv.org. https://arxiv.org/pdf/1703.03864.pdf.


view repo

Super-Pineapple

Project for Madhacks Fall 2017


view repo