Research map / Vision-language and multimodal / Cross-modal learning
M²VAE: Multi-Modal Multi-View Variational Autoencoder for Cold-start Item Recommendation
Chuan He, Yongchao Liu, Qiang Li, Chuntao Hong, Wenliang Zhong, Xin-Wei Yao
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 common view via PoE (ablation) | M2VAE · ML-20M | HR@5: 0.2673 -> 0.3045 | Improved |
| add disentangled contrastive loss (ablation) | M2VAE · ML-20M | NDCG@5: 0.4445 -> 0.4805 | Improved |
| add co-occurrence contrastive loss (ablation) | M2VAE · ML-20M | NDCG@5: 0.4390 -> 0.4805 | Improved |
| replace weighted PoE fusion with user-aware hierarchical MoE (ablation) | M2VAE · ML-20M | HR@5: 0.2650 -> 0.3045 | Improved |
| add self-gating units (ablation) | M2VAE · ML-20M | HR@5: 0.2801 -> 0.3045 | Improved |
| add modality-level MoE (ablation) | M2VAE · Video&Games | HR@5: 0.0312 -> 0.0367 | Improved |
| add view-level MoE (ablation) | M2VAE · Sports&Outdoors | HR@5: 0.0192 -> 0.0192 | No clear effect |
| generate common view in latent space rather than raw (early generate) (ablation) | M2VAE · Video&Games | NDCG@5: 0.0899 -> 0.1009 | Improved |
| use contrastive co-occurrence learning instead of pretraining | cold-start recommenders · ML-20M and Video&Games vs Sports&Outdoors | HR@5: CCFCRec 0.2969 vs GoRec 0.2908 (ML-20M); GoRec 0.0166 vs CCFCRec 0.0107 (Sports) | Mixed |
| use M2VAE instead of best baseline | cold-start recommenders · Video&Games | HR@5: 0.0326 -> 0.0367 | Improved |
Closest papers by meaning
- Towards Unified Multi-Modal Personalization: Large Vision-Language Models for Generative Recommendation and Beyond (ICLR 2024 · similarity 0.89)
- DR-VAE: Debiased and Representation-enhanced Variational Autoencoder for Collaborative Recommendation (AAAI 2025 · similarity 0.87)
- Preference Aware Dual Contrastive Learning for Item Cold-Start Recommendation (AAAI 2024 · similarity 0.87)
- Temporally and Distributionally Robust Optimization for Cold-Start Recommendation (AAAI 2024 · similarity 0.86)
- Beyond Graph Convolution: Multimodal Recommendation with Topology-aware MLPs (AAAI 2025 · similarity 0.86)
- Harnessing Multimodal Large Language Models for Multimodal Sequential Recommendation (AAAI 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.
