AI Paper Atlas

Research map / Reinforcement learning / Visual and representation RL

Task-Agnostic Dynamics Priors for Deep Reinforcement Learning

Yilun Du, Karthic Narasimhan

ICML 2019 · Reinforcement learning › Visual and representation RL · approach: Latent-space world models · 22 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 SpatialNet rather than RCNet or ConvLSTM for noisy-input predictionPhysics prediction · Physics prediction, ε=0.5MSE: 0.0268 (RCNet), 0.0072 (ConvLSTM) -> 0.0062Improved
Use SpatialNet for generalization to smaller objects moving fasterVideo prediction · Unseen smaller and faster objectsMSE: 0.0115 (RCNet), 0.0067 (ConvLSTM) -> 0.0039Improved
Use SpatialNet for prediction at a larger input resolutionVideo prediction · 168x168 input imagesMSE: 0.0060 -> 0.0042Improved
Initialize SpatialNet with physics-dataset pretrainingAtari video prediction · AssaultMSE: 0.00522 -> 0.00477Improved
Initialize SpatialNet with physics-dataset pretrainingAtari video prediction · DemonAttackMSE: 0.00562 -> 0.00433Improved
Initialize SpatialNet with physics-dataset pretrainingAtari video prediction · EnduroMSE: 0.00411 -> 0.00576Hurt
Initialize SpatialNet with physics-dataset pretrainingAtari video prediction · PongMSE: 0.00584 -> 0.00636Hurt
Initialize SpatialNet with physics-dataset pretrainingAtari video prediction · 10 Atari environmentsMSE: 12.9% decrease in MSE errorMixed
Use D2A instead of PPOAtari policy learning · Stochastic Atari environmentsAtari scoreMixed
Use D2A instead of PPOAtari policy learning · AsteroidsAtari score: 1321.0 ± 233.5 -> 2098.4 ± 102.0Improved
Use D2A instead of PPOAtari policy learning · BoxingAtari score: 54.3 ± 2.5 -> 8.9 ± 20.0Hurt

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