SOTAVerified

Proximal Policy Optimization Algorithms

2017-07-20Code Available2· sign in to hype

John Schulman, Filip Wolski, Prafulla Dhariwal, Alec Radford, Oleg Klimov

arXivPDF

Code Available — Be the first to reproduce this paper.

Reproduce

Code

Abstract

We propose a new family of policy gradient methods for reinforcement learning, which alternate between sampling data through interaction with the environment, and optimizing a "surrogate" objective function using stochastic gradient ascent. Whereas standard policy gradient methods perform one gradient update per data sample, we propose a novel objective function that enables multiple epochs of minibatch updates. The new methods, which we call proximal policy optimization (PPO), have some of the benefits of trust region policy optimization (TRPO), but they are much simpler to implement, more general, and have better sample complexity (empirically). Our experiments test PPO on a collection of benchmark tasks, including simulated robotic locomotion and Atari game playing, and we show that PPO outperforms other online policy gradient methods, and overall strikes a favorable balance between sample complexity, simplicity, and wall-time.

Tasks

Continuous ControlDota 2Multi-Goal Reinforcement LearningNeural Architecture SearchOpenAI GymPolicy Gradient MethodsReinforcement LearningSpeech Enhancement

Benchmark Results

DatasetModelMetricClaimedVerifiedStatus
NATS-Bench Topology, CIFAR-10PPO (Schulman et al., 2017)Test Accuracy94.02—Unverified
NATS-Bench Topology, CIFAR-100PPO (Schulman et al., 2017)Test Accuracy71.68—Unverified
NATS-Bench Topology, ImageNet16-120PPO (Schulman et al., 2017)Test Accuracy44.95—Unverified

Reproductions