Research map / Learning theory and algorithms / Optimal transport
Differentiable Hierarchical Optimal Transport for Robust Multi-View Learning
Dixin Luo, Hongteng Xu, Lawrence Carin
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 |
|---|---|---|---|
| replace pairwise sliced Wasserstein with hierarchical OT (view clustering) | multi-view learning · Caltech20 (unsupervised) | classification accuracy (%): 60.85 -> 71.07 with K=3 | Improved |
| add autoencoder reconstruction regularizer R_S to HOT K=3 (ablation) | HOT K=3 · Handwritten (semi-supervised) | classification accuracy (%): 90.05 -> 91.07 | Improved |
| add autoencoder regularizer to HOT K=3 (ablation) | HOT K=3 · Caltech7 (semi-supervised) | classification accuracy (%): 88.29 -> 88.98 | Improved |
| use unaligned data with HOT K=3 AE instead of aligned AECCA | HOT K=3 AE · Handwritten (semi-supervised) | classification accuracy (%): 88.53 -> 91.07 | Improved |
| use HOT K=3 instead of AECCA in unsupervised mode | HOT K=3 · Caltech20 (unsupervised) | classification accuracy (%): 66.75 -> 71.07 | Improved |
| use K=3 view clusters in HOT | HOT · Caltech7 (unsupervised) | classification accuracy (%): 82.89 -> 86.27 | Improved |
| use K=3 view clusters in HOT | HOT · Handwritten (unsupervised) | classification accuracy (%): 67.65 -> 87.12 | Improved |
| add autoencoder regularizer to HOT (no K clusters) (ablation) | HOT · Caltech20 (unsupervised) | classification accuracy (%): 60.85 -> 67.41 | Improved |
| add autoencoder regularizer to HOT K=3 (ablation) | HOT K=3 · Caltech20 (unsupervised) | classification accuracy (%): 71.07 -> 69.48 | Hurt |
| remove HOG or LBP view (ablation) | HOT model · Caltech7 | classification accuracy (%) | Hurt |
| remove Gabor or CENTRIST view (ablation) | HOT model · Caltech7 | classification accuracy (%) | No clear effect |
Closest papers by meaning
- Hierarchical Refinement: Optimal Transport to Infinity and Beyond (ICML 2025 · similarity 0.86)
- Differentiable Generalized Sliced Wasserstein Plans (NeurIPS 2025 · similarity 0.85)
- Robust Multi-view Clustering against Imperfect Information (NeurIPS 2026 · similarity 0.85)
- Hierarchical Multi-Marginal Optimal Transport for Network Alignment (AAAI 2024 · similarity 0.85)
- PROTOCOL: Partial Optimal Transport-enhanced Contrastive Learning for Imbalanced Multi-view Clustering (ICML 2025 · similarity 0.85)
- Contrasting Multiple Representations with the Multi-Marginal Matching Gap (ICML 2024 · similarity 0.85)
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.
