AI Paper Atlas

Research map / Reinforcement learning / Policy gradient and actor-critic

Regularizing Trajectory Optimization with Denoising Autoencoders

Rinu Boney, Norman Di Palo, Mathias Berglund, Alexander Ilin, Juho Kannala, Antti Rasmus, Harri Valpola

NeurIPS 2019 · Reinforcement learning › Policy gradient and actor-critic · approach: Alternative policy update rules · 6 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
Add DAE regularization to CEM trajectory optimizationPETS · Half-cheetahAverage Return: 10955 ± 2865 -> 12967 ± 3216Improved
Use Adam with DAE regularization instead of unregularized AdamPETS · Half-cheetahAverage ReturnImproved
Use DAE regularization during open-loop trajectory optimizationPETS with Adam · Half-cheetahagreement between predicted and true rewardsImproved
Add DAE regularization to trajectory optimizationModel-based RL · Cartpolelearning speedNo clear effect
Train the dynamics model for more epochs after each episode (ablation)Single-model end-to-end training with CEM · Motor-control tasksearly learning performanceImproved
Add DAE regularization (ablation)Single-model end-to-end training with CEM · Motor-control taskslearning progressImproved
Replace DAE regularization with diagonal-Gaussian regularization (ablation)Adam trajectory optimization · Half-cheetahlearning performanceHurt
Allow optimized actions to include random noise (ablation)DAE-regularized model-based RL · Motor-control tasksasymptotic performanceImproved
Stop gradient propagation through the denoising autoencoder (ablation)DAE-regularized trajectory optimization · Experimentsoptimization performanceNo clear effect
Use DAE regularization with Adam instead of unregularized AdamModel-based RL · Half-cheetah end-to-end trainingearly learning speedImproved
Use DAE regularization without adding explicit explorationModel-based RL · Half-cheetahasymptotic performanceHurt

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