Research map / Learning theory and algorithms / Clustering
MHCN: A Hyperbolic Neural Network Model for Multi-view Hierarchical Clustering
Fangfei Lin, Bing Bai, Yiwen Guo, Hao Chen, Yazhou Ren, Zenglin Xu
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 |
|---|---|---|---|
| one-stage hyperbolic multi-view training instead of two-stage CMHHC | MHCN · MNIST-USPS, BDGP, Caltech, COIL-20, BBCSport, Multi-Fashion | Dendrogram Purity: 94.49% -> 99.22% on MNIST-USPS; 53.50% -> 78.93% on BBCSport | Improved |
| add multi-view alignment loss (ablation) | MHCN · MNIST-USPS | Dendrogram Purity: 41.29% -> 99.22% | Improved |
| add hyperbolic uniformity loss (ablation) | MHCN · MNIST-USPS | Dendrogram Purity: 19.42% -> 99.22% | Improved |
| add reconstruction loss (ablation) | MHCN · seven multi-view datasets | Dendrogram Purity: 99.07% -> 99.22% on MNIST-USPS; 75.17% -> 77.14% on Caltech | Improved |
| use hyperbolic autoencoders instead of Euclidean autoencoders (ablation) | MHCN · BDGP | Dendrogram Purity: 52.47% -> 96.22% | Improved |
| use hyperbolic autoencoders instead of Euclidean autoencoders (ablation) | MHCN · Caltech | Dendrogram Purity: 32.50% -> 77.14% | Improved |
| increase weight of hyperbolic uniformity loss from 0 to 1 (ablation) | MHCN · MNIST-USPS (2D Poincare ball) | tree balance (qualitative) | Improved |
| set alpha to 0 so alignment loss is dropped (ablation) | MHCN · seven datasets | Dendrogram Purity | Hurt |
| tune temperatures tau_align and tau_uni near 1.0 (ablation) | MHCN · seven datasets | Dendrogram Purity | Mixed |
| use mini-batch one-stage MHCN instead of agglomerative HAC | MHCN · NR-MNIST | running time (s): 1823.43s (HAC-Single) -> 900.92s | Improved |
| apply MHCN inductively to unseen data | MHCN · NR-MNIST | Dendrogram Purity: 98.71% -> 96.54% | No clear effect |
Closest papers by meaning
- Wasserstein-Aligned Hyperbolic Multi-View Clustering (AAAI 2026 · similarity 0.91)
- Reciprocal Multi-Layer Subspace Learning for Multi-View Clustering (ICCV 2019 · similarity 0.88)
- Asymmetric Multi-View Clustering with Hyperbolic Uncertainty Modeling (ICML 2026 · similarity 0.88)
- Fully Hyperbolic Convolutional Neural Networks for Computer Vision (ICLR 2024 · similarity 0.86)
- Medusa: A Multi-Scale High-order Contrastive Dual-Diffusion Approach for Multi-View Clustering (CVPR 2025 · similarity 0.85)
- Improving Visual Recognition with Hyperbolical Visual Hierarchy Mapping (CVPR 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.
