Research map / Architectures and efficiency / RNNs and normalization
On Blackbox Backpropagation and Jacobian Sensing
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 |
|---|---|---|---|
| use GreedyColoring on intersection graph to choose perturbations | rainbow-LP · Sparse Jacobian, p=0.1 | chi/Delta | Helped |
| increase density from p=0.1 to p=0.3 | rainbow-LP · Sparse Jacobian recovery | measurements k (m=50): 30 -> 70 | Hurt |
| increase density from p=0.1 to p=0.3 | rainbow-LP · Sparse Jacobian recovery | chi/Delta (m=50): 0.32 -> 0.8625 | Hurt |
| use power-law alpha-model instead of p-model | rainbow-LP · Sparse Jacobian recovery | relative error (m=50): 0.1287 -> 0.0443 | Helped |
| alpha 0.5 vs 0.7 | rainbow-LP · Sparse Jacobian recovery (m=50) | relative error: 0.0443 -> 0.0326 | Helped |
| use rainbow-LP vs linear regression and finite differencing at low noise | rainbow-LP · MNIST convnet blackbox layer, std 9e-5 | validation error | No clear effect |
| use rainbow-LP at higher noise (std 0.008) | rainbow-LP · MNIST convnet blackbox layer | validation error | Helped |
| vary number of Jacobian measurements at std 0.1 (ablation) | rainbow-LP · MNIST convnet blackbox layer | validation error | Mixed |
| add sparsity and symmetry priors with ADMM instead of l2 linear regression (ablation) | rainbow ADMM · 50-link manipulator Jacobian | recovery accuracy | Helped |
| use Circulant perturbation matrix instead of random (ablation) | rainbow ADMM · 50-link manipulator Jacobian | recovery accuracy | No clear effect |
Closest papers by meaning
- Characterizing control between interacting subsystems with deep Jacobian estimation (NeurIPS 2025 · similarity 0.84)
- A Riemannian Framework for Learning Reduced-order Lagrangian Dynamics (ICLR 2025 · similarity 0.83)
- Improving equilibrium propagation without weight symmetry through Jacobian homeostasis (ICLR 2024 · similarity 0.83)
- Posterior Inference in Latent Space for Scalable Constrained Black-box Optimization (NeurIPS 2026 · similarity 0.83)
- Deep-Koopman-KANDy: Dictionary Discovery for Deep-Koopman Operators with Kolmogorov-Arnold Networks for Dynamics (NeurIPS 2026 · similarity 0.83)
- Straight-Through Meets Sparse Recovery: the Support Exploration Algorithm (ICML 2024 · similarity 0.83)
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.