AI Paper Atlas

Research map / Reinforcement learning / Visual and representation RL

RLlib Flow: Distributed Reinforcement Learning is a Dataflow Problem

Eric Liang, Zhanghao Wu, Michael Luo, Sven Mika, Joseph E. Gonzalez, Ion Stoica

NeurIPS 2021 · Reinforcement learning › Visual and representation RL · approach: Scalable RL systems and environments · 4 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
port to RLlib Flow dataflow model (minimal count)RLlib IMPALA · IMPALAlines of code: 694 -> 89Improved
port to RLlib Flow (conservative count incl. shared operators)RLlib IMPALA · IMPALAlines of code: 694 -> 362Improved
port to RLlib FlowRLlib A3C · A3Clines of code: 87 -> 11Improved
port to RLlib FlowRLlib A2C · A2Clines of code: 154 -> 25Improved
port to RLlib FlowRLlib DQN · DQNlines of code: 239 -> 87Improved
port to RLlib FlowRLlib PPO · PPOlines of code: 386 -> 79Improved
port to RLlib FlowRLlib Ape-X · Ape-Xlines of code: 250 -> 126Improved
port to RLlib Floworiginal MAML · MAMLlines of code: 370 -> 139Improved
count shared operators in RLlib Flow lines of codeRLlib Flow · Ape-Xcode reduction ratio: 1.9 -> 1.1Hurt
implement PPO in RLlib Flow instead of Spark StreamingPPO · CartPole-v0 distributedtraining time: up to 2.9xImproved
use RLlib Flow instead of original RLlib implementationRLlib · Atari throughput, IMPALA, CartPolethroughput / sample efficiencyNo clear effect

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy