Research map / Optimization / Implicit bias and generalization
On the Equivalence between Neural Network and Support Vector Machine
Yilan Chen, Wei Huang, Lam M. Nguyen, Tsui-Wei Weng
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 changed | Where | Measured | Effect |
|---|---|---|---|
| increase NN hidden width from 10^3 to 10^6 (ablation) | two-layer ReLU NN · MNIST 0/1 (100 test) | robustness certificate (x1e-3): 7.4485 -> 0.31539 | Hurt |
| increase NN hidden width from 10^3 to 10^6 (ablation) | two-layer ReLU NN · MNIST 0/1 (full test) | robustness certificate (x1e-3): 7.2708 -> 0.30997 | Hurt |
| replace finite-width NN (10^6) with equivalent infinite-width SVM certificate | NN robustness certification · MNIST 0/1 (100 test) | robustness certificate (x1e-3): 0.31539 -> 8.0541 | Improved |
| use KRR with l2 regularization lambda=1000 instead of lambda=0 (ablation) | KRR with NTK · MNIST 0/1 | robustness certificate: 790.06X | Improved |
| use KRR with l2 regularization lambda=100 instead of lambda=0 (ablation) | KRR with NTK · MNIST 0/1 | robustness certificate: 619.50X | Improved |
| use KRR with l2 regularization lambda=10 instead of lambda=0 (ablation) | KRR with NTK · MNIST 0/1 | robustness certificate: 156.90X | Improved |
| use KRR with l2 regularization lambda=0.001 instead of lambda=0 (ablation) | KRR with NTK · MNIST 0/1 | robustness certificate: 1.14X | Improved |
| increase KRR lambda from 0 to 1000 (ablation) | KRR with NTK · MNIST 0/1 | test accuracy: 99.95 -> 99.76 | Hurt |
| use soft margin SVM (lambda 0.064) instead of lambda=0 KRR | SVM with NTK · MNIST 0/1 | robustness certificate: 243.91X | Improved |
| use wider NN (10K vs narrower) when comparing training dynamics to SVM (ablation) | NN trained with soft margin loss · MNIST 0/1 | agreement of output dynamics with SVM | Improved |
Closest papers by meaning
- Divergence of Neural Tangent Kernel in Classification Problems (ICLR 2025 · similarity 0.87)
- Convergence Analysis of Newton's Method for Neural Networks in the Overparameterized Limit (NeurIPS 2026 · similarity 0.86)
- Neural Tangent Kernels Motivate Cross-Covariance Graphs in Neural Networks (ICML 2024 · similarity 0.86)
- Adaptive kernel predictors from feature-learning infinite limits of neural networks (ICML 2025 · similarity 0.86)
- An Infinite-Width Analysis on the Jacobian-Regularised Training of a Neural Network (ICML 2024 · similarity 0.86)
- On the Impacts of the Random Initialization in the Neural Tangent Kernel Theory (NeurIPS 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.
