AI Paper Atlas

Research map / Probabilistic and scientific modeling / Dynamical systems

Sparsely Changing Latent States for Prediction and Planning in Partially Observable Domains

Christian Gumbsch, Martin V. Butz, Georg Martius

NeurIPS 2021 · Probabilistic and scientific modeling › Dynamical systems · approach: Latent dynamics models · 1 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 GateL0RD with latent-state L0 regularization (λ=0.001) instead of standard RNNsBilliard Ball · Billiard Ballautoregressive prediction errorImproved
Use GateL0RD (λ=0.001) instead of standard RNNsRobot Remote Control · Robot Remote Control test setprediction errorImproved
Use GateL0RD (λ=0.001) instead of standard RNNsRobot Remote Control · Robot Remote Control generalization setprediction errorImproved
Use GateL0RD instead of standard RNNsRobot Remote Control MPC · Robot Remote Controlsuccess rateImproved
Use GateL0RD (λ=0.0001) instead of standard RNNsShepherd · Shepherd 100-step predictionprediction errorImproved
Use GateL0RD instead of standard RNNsShepherd MPC · Shepherdsuccess rateImproved
Replace the LSTM memory module with GateL0RD (λ=0.01)PPO actor-critic · MiniGrid taskssuccess rateImproved
Replace the LSTM memory module with GateL0RD (λ=0.01)PPO actor-critic · MiniGrid taskssample efficiencyImproved
Set λ=0.001 rather than use stronger regularization (ablation)GateL0RD · Billiard Ballprediction errorImproved
Increase regularization strength to λ=0.1 (ablation)GateL0RD · Billiard Ballprediction errorHurt
Vary regularization strength within λ∈[0, 0.01] (ablation)GateL0RD · Billiard Ballprediction errorNo clear effect
Remove latent-state regularization (λ=0) (ablation)GateL0RD · Fetch Pick&Place with continuously changing dynamicsprediction performanceImproved

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