Research map / Learning theory and algorithms / Kernels and regression theory
Generalization Analysis for Contrastive Representation Learning
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 two vector-contraction steps to bound the ℓ2-Lipschitz contrastive loss | Contrastive representation learning · Contrastive representation learning | generalization bound: improves the bounds in [cite] by a factor of \sqrt{k} | Helped |
| Use covering numbers and an expanded set of negative-example triples for the ℓ∞-Lipschitz loss | Contrastive representation learning · Contrastive representation learning | generalization bound: improves the bounds in [cite] by a factor of k | Helped |
| Use a self-bounding Lipschitz condition and an optimistic bound | Contrastive representation learning · Contrastive representation learning | generalization bound | Helped |
| Preserve coupling among representation features in the Rademacher complexity analysis | Linear and nonlinear representations · Representation learning | feature-dimension dependency in the bound: the factor of d in Eq. [ref] is replaced by √d | Helped |
| Use the ℓ∞-Lipschitz analysis rather than the ℓ2-Lipschitz analysis for downstream classification | Downstream mean classifier · Downstream classification | negative-example dependency in the bound: logarithmic dependency on k | Helped |
| Apply the self-bounding bound in an interpolation setting with zero empirical error | Contrastive representation learning · Interpolation setting | generalization bound rate: \widetilde{O}\big(Bn^{-1}+G_s^2R^4n^{-1}\big) | Helped |
Closest papers by meaning
- Generalization Analysis for Deep Contrastive Representation Learning (AAAI 2025 · similarity 0.93)
- Generalization Analysis for Supervised Contrastive Representation Learning under Non-IID Settings (ICML 2025 · similarity 0.92)
- Generalizability of Neural Networks Minimizing Empirical Risk Based on Expressive Power (ICLR 2025 · similarity 0.90)
- Towards Generalization beyond Pointwise Learning: A Unified Information-theoretic Perspective (ICML 2024 · similarity 0.89)
- Optimal Sample Complexity of Contrastive Learning (ICLR 2024 · similarity 0.89)
- Fantastic Generalization Measures are Nowhere to be Found (ICLR 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.