Research map / Learning theory and algorithms / Kernels and regression theory
Deep Semi-Random Features for Nonlinear Function Approximation
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 |
|---|---|---|---|
| adapt random features by learning unit weights (LSR) | semi-random network · letter | error (%): 14.9 -> 6.5 | Helped |
| adapt random features by learning unit weights (LSR) | semi-random network · webspam | error (%): 1.1 -> 1.1 | No clear effect |
| adapt random features by learning unit weights (LSR) | semi-random network · covtype | error (%): 20.2 -> 5.7 | Helped |
| increase LSR units from baseline to 2× (ablation) | LSR · MNIST | test error (%): 0.97 -> 0.78 | Helped |
| increase LSR units from 2× to 4× (ablation) | LSR · CIFAR10 | test error (%): 17.4 -> 18.7 | Hurt |
| use implicit-ensemble units instead of LSR (ablation) | semi-random network · MNIST | test error (%): 0.97 -> 0.59 | Helped |
| increase LSR-IE units from 2× to 4× (ablation) | LSR-IE · MNIST | test error (%): 0.47 -> 0.54 | Hurt |
| increase LSR-IE units from baseline to 4× (ablation) | LSR-IE · CIFAR10 | test error (%): 20.0 -> 14.9 | Helped |
| increase LSR-IE units from baseline to 4× (ablation) | LSR-IE · SVHN | test error (%): 6.9 -> 4.8 | Helped |
| increase random-feature units from baseline to 16× (ablation) | RF · MNIST | test error (%): 8.80 -> 2.69 | Helped |
| use deeper rather than wider LSR networks at the same total unit count (ablation) | LSR · UCI datasets | test error | Helped |
Closest papers by meaning
- Random deep neural networks are biased towards simple functions (NeurIPS 2019 · similarity 0.87)
- Generalization Bounds of Stochastic Gradient Descent for Wide and Deep Neural Networks (NeurIPS 2019 · similarity 0.86)
- Friendly Sharpness-Aware Minimization (CVPR 2024 · similarity 0.86)
- Adaptive Random Feature Regularization on Fine-tuning Deep Neural Networks (CVPR 2024 · similarity 0.85)
- Asymptotics of feature learning in two-layer networks after one gradient-step (ICML 2024 · similarity 0.85)
- On the Minimal Degree Bias in Generalization on the Unseen for non-Boolean Functions (ICML 2024 · 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.