Research map / Label-efficient and robust learning / Self-supervised and contrastive
Learning about an exponential amount of conditional distributions
Mohamed Belghazi, Maxime Oquab, David Lopez-Paz
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 |
|---|---|---|---|
| use adversarial training instead of moment matching (ablation) | NC · 3D Gaussian, a=(1, 0, 0), r=(0, 1, 1) | conditional parameter error: 0.67 -> 0.13 | Improved |
| condition both the discriminator and NC on available and requested masks (ablation) | NC · 3D Gaussian | Euclidean parameter error: 0.12 -> 0.07 | Improved |
| remove mask conditioning from the discriminator while retaining it in NC (ablation) | NC · 3D Gaussian | Euclidean parameter error: 0.07 -> 0.17 | Hurt |
| remove mask conditioning from NC while retaining it in the discriminator (ablation) | NC · 3D Gaussian | Euclidean parameter error: 0.07 -> 0.15 | Hurt |
| use NC features rather than the strongest listed baseline features | SVHN classification with linear SVM · SVHN test | test error: 16.61 vs 17.12 | Mixed |
| use NC features rather than VAEAC features | CelebA attribute classification · CelebA test | balanced accuracy (mean): 66.06 -> 82.21 | Improved |
Closest papers by meaning
- Neural Conditional Probability for Uncertainty Quantification (NeurIPS 2024 · similarity 0.85)
- Generative Conditional Distributions by Neural (Entropic) Optimal Transport (ICML 2024 · similarity 0.83)
- A Likelihood Based Approach to Distribution Regression Using Conditional Deep Generative Models (ICML 2025 · similarity 0.82)
- Wasserstein Gradient Boosting: A Framework for Distribution-Valued Supervised Learning (NeurIPS 2024 · similarity 0.82)
- Beyond Textual Constraints: Learning Novel Diffusion Conditions with Fewer Examples (CVPR 2024 · similarity 0.81)
- Zero-Shot Conditioning of Score-Based Diffusion Models by Neuro-Symbolic Constraints (AAAI 2025 · similarity 0.81)
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.
