Research map / Probabilistic and scientific modeling / Dynamical systems
The least-control principle for local learning at equilibrium
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 changed | Where | Measured | Effect |
|---|---|---|---|
| LCP with dynamic inversion vs recurrent backprop | equilibrium RNN · MNIST test | accuracy (%): 97.67 -> 97.55 | No clear effect |
| replace direct linear feedback with dynamic inversion controller (ablation) | LCP equilibrium RNN · MNIST test | accuracy (%): 97.45 -> 97.55 | Helped |
| learn feedback weights (Kolen-Pollack) instead of tying S=W^T (ablation) | LCP-DI · MNIST test | accuracy (%): 97.55 -> 97.47 | No clear effect |
| LCP-DI vs RBP on convolutional equilibrium RNN | CIFAR-10 equilibrium RNN · CIFAR-10 test | accuracy (%): 80.14 -> 80.26 | No clear effect |
| LCP-DI vs RBP for implicit neural representation | equilibrium RNN · INR image dataset | PSNR (dB): 25.47 -> 25.11 | No clear effect |
| LCP meta-learning vs other first-order methods | Omniglot meta-learning · Omniglot 20-way 1-shot | accuracy (%): 89.72 -> 91.00 | Helped |
| LCP vs second-order iMAML | Omniglot meta-learning · Omniglot 20-way 1-shot | accuracy (%): 94.46 -> 91.00 | Hurt |
| use derivative-of-nonlinearity-gated updates when column space condition is not met (ablation) | LCP-LF · MNIST | accuracy (%) | Helped |
Closest papers by meaning
- Minimizing Control for Credit Assignment with Strong Feedback (ICML 2022 · similarity 0.88)
- Feedback control guides credit assignment in recurrent neural networks (NeurIPS 2024 · similarity 0.87)
- Dynamics of Supervised and Reinforcement Learning in the Non-Linear Perceptron (NeurIPS 2024 · similarity 0.86)
- How Deep Do We Need: Accelerating Training and Inference of Neural ODEs via Control Perspective (ICML 2024 · similarity 0.86)
- Learning Dynamics of RNNs in Closed-Loop Environments (NeurIPS 2025 · similarity 0.85)
- Contractive Dynamical Imitation Policies for Efficient Out-of-Sample Recovery (ICLR 2025 · similarity 0.85)
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.