Research map / Label-efficient and robust learning / Recognition and long-tail
Robust Multi-View Learning via Representation Fusion of Sample-Level Attention and Alignment of Simulated Perturbation
Jie Xu, Na Zhao, Gang Niu, Masashi Sugiyama, Xiaofeng 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 |
|---|---|---|---|
| add sample-level attention, noise and unusable perturbation to base model (ablation) | RML · YoutubeVideo (noisy-label classification, 50% noise) | ACC: 0.307 -> 0.598 | Improved |
| add sample-level attention, noise and unusable perturbation to base model (ablation) | RML · BDGP (noisy-label classification, 50% noise) | ACC: 0.864 -> 0.936 | Improved |
| add sample-level attention alone to base model (ablation) | RML · BDGP (noisy-label classification, 50% noise) | ACC: 0.864 -> 0.698 | Hurt |
| add sample-level attention alone to base model (ablation) | RML · YoutubeVideo (noisy-label classification, 50% noise) | ACC: 0.307 -> 0.337 | Improved |
| add noise perturbation to attention plus unusable perturbation (ablation) | RML · BDGP (noisy-label classification, 50% noise) | ACC: 0.834 -> 0.936 | Improved |
| add unusable perturbation to attention plus noise perturbation (ablation) | RML · YoutubeVideo (noisy-label classification, 50% noise) | ACC: 0.421 -> 0.598 | Improved |
| add unusable perturbation to attention plus noise perturbation (ablation) | RML · BDGP (noisy-label classification, 50% noise) | ACC: 0.957 -> 0.936 | Hurt |
| add sample-level attention to the two perturbations (ablation) | RML · YoutubeVideo (noisy-label classification, 50% noise) | ACC: 0.581 -> 0.598 | Improved |
| add sample-level attention to the two perturbations (ablation) | RML · DHA (noisy-label classification, 50% noise) | ACC: 0.678 -> 0.610 | Hurt |
| add RML regularization to transformer with cross-entropy | Trans.+L_CE · BDGP (noisy-label classification, 50% noise) | ACC: 0.437 -> 0.936 | Improved |
| replace L_CE with L_MCE under RML (ablation) | RML · DHA (noisy-label classification, 50% noise) | ACC: 0.608 -> 0.610 | No clear effect |
| add RML regularization to UCCH | UCCH · NUS-WIDE (Image->Text, 16 bits) | MAP: 0.698 -> 0.733 | Improved |
Closest papers by meaning
- CFDM: Contrastive Fusion and Disambiguation for Multi-View Partial-Label Learning (AAAI 2025 · similarity 0.89)
- Multi-View Representation Learning via Total Correlation Objective (NeurIPS 2021 · similarity 0.89)
- Rethinking Multi-view Representation Learning via Distilled Disentangling (CVPR 2024 · similarity 0.88)
- ITO: Multi-View Alignment and Training-Time Fusion for Image-Text Pretraining (NeurIPS 2026 · similarity 0.87)
- SeqMvRL: A Sequential Fusion Framework for Multi-view Representation Learning (CVPR 2025 · similarity 0.87)
- Deep Incomplete Multi-view Learning via Cyclic Permutation of VAEs (ICLR 2025 · similarity 0.86)
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.
