Research map / Learning theory and algorithms / Sample complexity
A Data-Augmented Contrastive Learning Approach to Nonparametric Density Estimation
Chenghao Li, Yuanyuan Lin
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 |
|---|---|---|---|
| add data augmentation (rho>0) to nonparametric noise contrastive estimation (ablation) | NNCE · Model 1 density, d=10 (mean of 5 runs) | Spearman rank correlation: 0.821 -> 0.875 | Improved |
| add data augmentation (rho>0) to nonparametric noise contrastive estimation (ablation) | NNCE · Model 1 density, d=8 (mean of 5 runs) | Spearman rank correlation: 0.884 -> 0.908 | Improved |
| add data augmentation to nonparametric noise contrastive estimation (ablation) | NNCE · 2-D density models | L2 distance | No clear effect |
| increase rho from 0 to 5 (ablation) | DANNCE · Model 1 density, d>=5 | Spearman rank correlation | Improved |
| increase rho (ablation) | DANNCE · 2-D models | L2 distance | No clear effect |
| increase rho (ablation) | DANNCE · Model 1 density, d=3 and 4 | Spearman rank correlation | No clear effect |
| use DANNCE instead of Roundtrip | Roundtrip · Model 1 (2-D) | L2 distance: .3865 -> .2129 | Improved |
| use DANNCE instead of telescoping density-ratio estimation | TRE · Model 3 (2-D) | L2 distance: 2.651 -> 2.463 | Improved |
| use DANNCE instead of telescoping density-ratio estimation | TRE · Model 1 density, d>=8 | Spearman rank correlation: 0.929 -> 0.875 | Hurt |
| use DANNCE instead of telescoping density-ratio estimation | TRE · Shuttle anomaly detection | precision at k: 0.914 -> 0.970 | Improved |
| use DANNCE instead of Roundtrip | Roundtrip · Mammography anomaly detection | precision at k: 0.482 -> 0.481 | No clear effect |
| use DANNCE instead of kernel-based two-stage estimator | Two-stage · Model 1 (2-D) | L2 distance: .6781 -> .2129 | Improved |
Closest papers by meaning
- Conditional Noise-Contrastive Estimation of Unnormalised Models (ICML 2018 · similarity 0.87)
- Nonparametric Estimation of a Factorizable Density using Diffusion Models (NeurIPS 2026 · similarity 0.86)
- Adaptive Multi-stage Density Ratio Estimation for Learning Latent Space Energy-based Model (NeurIPS 2022 · similarity 0.86)
- A Unified View on Learning Unnormalized Distributions via Noise-Contrastive Estimation (ICML 2025 · similarity 0.85)
- Density Ratio Estimation with Conditional Probability Paths (ICML 2025 · similarity 0.84)
- Density-Softmax: Efficient Test-time Model for Uncertainty Estimation and Robustness under Distribution Shifts (ICML 2024 · similarity 0.84)
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.
