AI Paper Atlas

Research map / Reinforcement learning / Visual and representation RL

Shaping Belief States with Generative Environment Models for RL

Karol Gregor, Danilo Jimenez Rezende, Frederic Besse, Yan Wu, Hamza Merzic, Aaron van den Oord

NeurIPS 2019 · Reinforcement learning › Visual and representation RL · approach: Learned state and action representations · 51 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
Increase overshoot length (ablation)SimCore agents (all losses) · Random Cityposition and map decoding accuracyImproved
Use generative prediction loss instead of contrastive for map decoding (ablation)SimCore · Random Citymap decoding MSE: 160 -> 117Improved
Use contrastive loss instead of generative loss for position decoding (ablation)SimCore · Random Cityposition decoding accuracyImproved
Use contrastive loss for mapping (ablation)Contrastive SimCore · Random Citymap decodingHurt
Add Kanerva memory to generative model (ablation)Generative SimCore · Random Citymap decoding accuracyImproved
Use Kanerva memory instead of pure LSTM or slot-based memory (ablation)LSTM belief state · Random Citymap decodingImproved
Add generative SimCore model loss to LSTM agentLSTM IMPALA agent · DeepMind LabRL score / data efficiencyImproved
Add generative SimCore model loss to RMA agentRMA agent · DeepMind LabRL score / data efficiencyImproved
Use longer overshoot lengths (ablation)SimCore agent · DeepMind LabRL performanceNo clear effect
Use Kanerva memory in the RL setting (ablation)SimCore agent · DeepMind LabRL training stabilityHurt
Add SimCore to LSTM agentLSTM agent · Voxel levels (Food, HighFood, Cliff, Bridge)RL scoreImproved
Add SimCore to LSTM agentLSTM agent · Voxel levelssensitivity to Adam hyper-parameters and unroll lengthImproved

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