Research map / Label-efficient and robust learning / Recognition and long-tail
Multi-view Learning via Trusted Pairwise Entity Energy
Yalan Qin, Guorui Feng, Xinpeng Zhang
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 ENIG evidential fusion versus TLC baseline (single best view) | long-tailed multi-view classifier · Scene15-LT | accuracy: 60.5 -> 70.0 | Improved |
| use ENIG versus TLC (single best view) | long-tailed multi-view classifier · Handwritten-LT | accuracy: 95.2 -> 97.0 | Improved |
| use ENIG versus best baselines on Caltech101-LT (single view) | long-tailed multi-view classifier · Caltech101-LT | accuracy: 88.9 -> 89.0 | No clear effect |
| use uncertainty-based multi-view fusion (ENIG) versus concatenation baselines | long-tailed multi-view classifier · Handwritten-LT (concatenated views) | accuracy: 98.4 -> 98.4 | No clear effect |
| use all views (concatenated) instead of best single view | ENIG · Scene15-LT | accuracy: 70.0 -> 74.2 | Improved |
| add ENIG distribution (ablation) | proposed method · CUB-LT / Scene15-LT / HMDB-LT (random view) | accuracy: 0.840, 0.680, 0.625 without ENIG | Improved |
| add both L_ace and KL regularizer (full objective) versus one term (ablation) | proposed method · PIE-LT | accuracy: 0.790 -> 0.910 | Improved |
| add both loss terms (full objective) versus one term (ablation) | proposed method · HMDB-LT | accuracy: 0.546 -> 0.710 | Improved |
| lambda_l set to 0.1 among [0.0001 to 1] (ablation) | proposed method · PIE-LT | accuracy | Improved |
| evaluate accuracy on head, medium and tail classes | proposed method · long-tailed multi-view datasets | accuracy | Mixed |
Closest papers by meaning
- Enhancing Testing-Time Robustness for Trusted Multi-View Classification in the Wild (CVPR 2025 · similarity 0.86)
- Navigating Conflicting Views: Harnessing Trust for Learning (ICML 2025 · similarity 0.86)
- Trusted Multi-View Classification with Expert Knowledge Constraints (ICML 2025 · similarity 0.85)
- Deep Fuzzy Multi-view Learning for Reliable Classification (ICML 2025 · similarity 0.85)
- Enhancing Multi-View Classification Reliability with Adaptive Rejection (AAAI 2025 · similarity 0.84)
- Towards Heterogeneous Long-tailed Learning: Benchmarking, Metrics, and Toolbox (NeurIPS 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.
