Research map / Probabilistic and scientific modeling / Variational and generative sampling
Population Matching Discrepancy and Applications in Deep Learning
Jianfei Chen, Chongxuan LI, Yizhong Ru, Jun Zhu
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 PMD instead of MMD | domain regularizer · Office a->w | accuracy: 76.9 -> 86.2 | Improved |
| use PMD instead of MMD | domain regularizer · Office a->d | accuracy: 78.4 -> 82.7 | Improved |
| use PMD instead of MMD | domain regularizer · Office d->a | accuracy: 64.9 -> 64.3 | No clear effect |
| use PMD instead of MMD | domain regularizer · Office w->a | accuracy: 68.1 -> 66.8 | No clear effect |
| use PMD instead of MMD | domain regularizer · Office average | accuracy: 80.7 -> 82.6 | Improved |
| add PMD domain regularization to source-only NN | NN · Office average | accuracy: 75.7 -> 82.6 | Improved |
| use PMD instead of MMD | domain regularizer · Office d->a | convergence speed | Improved |
| decrease mini-batch size |B| in PMD (ablation) | PMD · MNIST fc generative | final PMD | Improved |
| vary population size N and use exact vs approximated matching (ablation) | PMD · MNIST fc generative | final PMD | No clear effect |
| use PMD instead of MMD for generation on pixel space | generative model · SVHN and LFW | image quality | Improved |
| increase population size N for MMD (ablation) | MMD · MNIST fc generative | convergence | Mixed |
| increase population size N (ablation) | PMD · toy Gaussians | relative approximation error: proportional to N^-0.23 | Improved |
Closest papers by meaning
- Generative Models and Model Criticism via Optimized Maximum Mean Discrepancy (ICLR 2017 · similarity 0.87)
- Central Moment Discrepancy (CMD) for Domain-Invariant Representation Learning (ICLR 2017 · similarity 0.87)
- Dataset Distillation with Neural Characteristic Function: A Minmax Perspective (CVPR 2025 · similarity 0.85)
- Efficient Discrepancy Testing for Learning with Distribution Shift (NeurIPS 2024 · similarity 0.85)
- Sobolev Regularized MMD Gradient Flow (NeurIPS 2026 · similarity 0.85)
- Expressive Score-Based Priors for Distribution Matching with Geometry-Preserving Regularization (ICML 2025 · 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.
