AI Paper Atlas

Research map / Architectures and efficiency / RNNs and normalization

Model-Based Control with Sparse Neural Dynamics

Ziang Liu, Genggeng Zhou, Jeff He, Tobia Marcucci, Li Fei-Fei, Jiajun Wu, Yunzhu Li

NeurIPS 2023 · Architectures and efficiency › RNNs and normalization · approach: Differentiable optimization layers · 0 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 MIP-based planning instead of MPPI with the learned dynamics modelSparse Dynamics · Object pushing, object sorting, rope manipulationopen-loop planning performanceImproved
Use sparsified dynamics models with MIP-based control rather than the evaluated RL baselinesClosed-loop control · CartPole-v1, Reacher-v4closed-loop control performanceImproved
Reduce the number of ReLUs while retaining a reasonably accurate dynamics modelSparse Dynamics · Object pushing, object sorting, rope manipulation, CartPole-v1, Reacher-v4closed-loop control performanceImproved
Reduce the number of ReLUs in the dynamics modelSparse Dynamics · Object pushing, object sorting, rope manipulationlong-horizon prediction performanceNo clear effect
Replace nonlinear units with identity mappings as well as removing unitsDynamics models at matched ReLU counts · Object pushing, object sorting, rope manipulationprediction errorImproved
Use the proposed nonlinearity sparsification instead of channel pruningObject pushing · Object pushingclosed-loop control performanceImproved
Sparsify the model to two ReLUsOverparameterized MLP for piecewise affine recovery · Piecewise affine functionsground-truth system recoveryImproved
Use channel pruning instead of the proposed sparsificationOverparameterized MLP for piecewise affine recovery · Piecewise affine functionsground-truth system recoveryHurt
Use a simplified learned dynamics model as the forward model for PPOPPO using learned dynamics · CartPole-v1closed-loop control 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