AI Paper Atlas

Research map / Graph learning / GNN architectures

Reward Propagation Using Graph Convolutional Networks

Martin Klissarov, Doina Precup

NeurIPS 2020 · Graph learning › GNN architectures · approach: Deep graph convolution and simplification · 2 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
Use GCN-propagated reward information as a potential for reward shapingPPO · MiniWorld-FourRooms-v0, MiniWorld-MyWayHome-v0, MiniWorld-MyWayHomeNoisyTv-v0scoreImproved
Use GCN-propagated reward information as a potential for reward shapingPPO · Atari 2600 gamesperformanceImproved
Use GCN-propagated reward information as a potential for reward shapingPPO · Delayed MuJoCo environmentsperformanceImproved
Use GCN message passing instead of the exact forward-backward message computationpotential-based reward shaping in FourRooms and FourRoomsTraps · FourRooms and FourRoomsTrapscumulative steps (regret)No clear effect
Replace the learned GCN potential with an L2 distance-to-goal potentialPPO reward shaping · MiniWorld environmentsperformanceNo clear effect
Use RND or ICM exploration shaping instead of the GCN potentialPPO · MiniWorld-MyWayHomeNoisyTv-v0performanceHurt
Increase α above 0.5, giving more weight to the regular return (ablation)Φ_GCN · MiniWorld-FourRooms-v0performanceImproved
Increase η, weighting propagation loss more strongly (ablation)Φ_GCN · SMaze-v0propagation complexityImproved

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