Research map / Learning theory and algorithms / Kernels and regression theory
Contrastive Learning Can Find An Optimal Basis For Approximately View-Invariant Functions
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 |
|---|---|---|---|
| Apply Kernel PCA to a learned positive-pair kernel | Contrastive learning models · Overlapping regions and MNIST synthetic tasks | eigenfunction alignment | Helped |
| Use different kernel parameterizations, including linear and hypersphere kernels | Contrastive learning models · Overlapping regions and MNIST synthetic tasks | eigenfunction alignment | No clear effect |
| Use different contrastive losses, including XEnt and spectral loss | Hypersphere-based models · Overlapping regions | eigenfunction alignment | No clear effect |
| Constrain the kernel approximation, such as reducing output dimension, fixing the linear-head norm, or fixing the hypersphere bias to zero (ablation) | Contrastive learning models · Overlapping regions and MNIST synthetic tasks | eigenfunction alignment | Hurt |
| Constrain the kernel approximation, such as reducing output dimension, fixing the linear-head norm, or fixing the hypersphere bias to zero (ablation) | Contrastive learning models · Overlapping regions and MNIST synthetic tasks | eigenvalue estimation accuracy | Hurt |
| Decrease augmentation strength by increasing the number of kept pixels (ablation) | MNIST synthetic task · MNIST synthetic task | number of recovered eigenfunctions | Hurt |
Closest papers by meaning
- When Can We Approximate Wide Contrastive Models with Neural Tangent Kernels and Principal Component Analysis? (AAAI 2025 · similarity 0.90)
- Self-Supervised Contrastive Learning is Approximately Supervised Contrastive Learning (NeurIPS 2025 · similarity 0.89)
- Generalization Analysis for Supervised Contrastive Representation Learning under Non-IID Settings (ICML 2025 · similarity 0.88)
- On the Similarities of Embeddings in Contrastive Learning (ICML 2025 · similarity 0.88)
- Contrastive Learning is Spectral Clustering on Similarity Graph (ICLR 2024 · similarity 0.88)
- Bridging Mini-Batch and Asymptotic Analysis in Contrastive Learning: From InfoNCE to Kernel-Based Losses (ICML 2024 · similarity 0.88)
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.