Research map / Label-efficient and robust learning / Recognition and long-tail
DIN: Dual Impulse Network for Multi-view Representation Learning
Yilin Wu, Weihong Lin, Renjie Lin, Zihan Fang, Shide Du, Shiping Wang
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 integration network on top of the two impulse networks (ablation) | DIN · ALOI | ACC: 87.10 -> 97.59 | Improved |
| add integration network on top of the two impulse networks (ablation) | DIN · UCI | ACC: 86.50 -> 95.58 | Improved |
| add integration network on top of the two impulse networks (ablation) | DIN · IAPR | ACC: 62.33 -> 69.84 | Improved |
| add integration network on top of the two impulse networks (ablation) | DIN · NUS-WIDE | ACC: 50.00 -> 52.01 | Improved |
| use channel-impulse network only instead of attribute-impulse network only (ablation) | DIN · Flickr vs WebKB-texas | ACC: Flickr 70.92 vs 56.70; WebKB-texas 64.88 vs 66.67 | Mixed |
| combine attribute and channel impulse networks (no integration) versus attribute only (ablation) | DIN · ALOI | ACC: 78.22 -> 87.10 | Improved |
| combine both impulse networks without integration versus attribute only (ablation) | DIN · Flickr | ACC: 70.92 -> 70.09 | Hurt |
| increase compression ratios alpha and beta to 4 or above (ablation) | DIN · MNIST | ACC | Hurt |
| use more GCN layers than two (ablation) | DIN · multiple datasets | ACC | Hurt |
| DIN versus best baseline on MNIST | DIN · MNIST (10% labels) | ACC: 93.84 -> 93.81 | No clear effect |
Closest papers by meaning
- Rethinking Multi-view Representation Learning via Distilled Disentangling (CVPR 2024 · similarity 0.85)
- Deep Embedded Complementary and Interactive Information for Multi-View Classification (AAAI 2020 · similarity 0.85)
- Incomplete Multi-view Sequential Representation Learning with Cascaded Cross-view Evolution Attention (TPAMI 2026 · similarity 0.85)
- Uncertainty-Aware Multi-View Representation Learning (AAAI 2021 · similarity 0.85)
- Understanding Latent Correlation-Based Multiview Learning and Self-Supervision: An Identifiability Perspective (ICLR 2022 · similarity 0.85)
- Learning 1D Causal Visual Representation with De-focus Attention Networks (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.
