AI Paper Atlas

Research map / Probabilistic and scientific modeling / Dynamical systems

Residual Pathway Priors for Soft Equivariance Constraints

Marc Anton Finzi, Gregory Benton, Andrew Gordon Wilson

NeurIPS 2021 · Probabilistic and scientific modeling › Dynamical systems · approach: State-space inference for stochastic systems · 3 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 a soft convolutional prior instead of an unconstrained MLPRPP on CIFAR-10 · CIFAR-10classification error: 37.61 ± 0.56 -> 12.62 ± 0.34Improved
Use a soft convolutional prior instead of a standard convolutional modelRPP on CIFAR-10 · CIFAR-10classification error: 12.03 ± 0.46 -> 12.62 ± 0.34No clear effect
Use a soft convolutional prior instead of an unconstrained MLPRPP on Energy · EnergyMSE: 0.39 ± 0.48 -> 0.73 ± 0.44Mixed
Use a soft convolutional prior instead of an unconstrained MLPRPP on Fertility · FertilityMSE: 0.049 ± 0.0044 -> 0.060 ± 0.0097Hurt
Use a soft convolutional prior instead of an unconstrained MLPRPP on Pendulum · PendulumMSE: 4.65 ± 0.50 -> 4.25 ± 0.50Improved
Use a soft convolutional prior instead of an unconstrained MLPRPP on Wine · WineMSE: 0.66 ± 0.058 -> 0.69 ± 0.031Hurt
Add the RPP equivariance prior to the transition modelMujoco transition models · Swimmer-v2 and Hopper-v2rollout relative error: Swimmer-v2: 3.9 ± 1.0 -> 2.75 ± \bf{0.31}; Hopper-v2: 9.8±0.5 -> 7.0±\bf{ 0.7}Improved
Add the RPP equivariance prior to the transition modelMujoco transition models · Ant-v2rollout relative error: 16.0\bf{±0.3} -> 20.0±1.1Hurt
Use RPP-EMLP for both policy and critic rather than policy alone (ablation)SAC on Mujoco · Mujoco locomotionaverage rewardImproved
Use RPP-EMLP for policy and criticSAC on Humanoid-v2 · Humanoid-v2training stabilityHurt
Use an equivariant architecture instead of RPP when the symmetry is approximateDynamical-system models · Modified inertia and windy pendulumtest performanceHurt
Use an equivariant architecture instead of RPP when the specified symmetry is wrongDynamical-system models · Misspecified inertia and double pendulumprediction 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