Research map / Image generation / Autoregressive and latent generation
Compatibility Family Learning for Item Recommendation and Generation
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 PCD with multiple compatible prototypes instead of a single L2 embedding | Fashion-MNIST+1+2 · Fashion-MNIST+1+2 | AUC: 0.8651 ± 0.0052 -> 0.8830 ± 0.0032 | Helped |
| Use PCD rather than Monomer | Amazon also-bought/viewed recommendation · Amazon also-bought/viewed average | error rate: 6.79% -> 6.26% | Helped |
| Use PCD rather than Monomer | Amazon also-viewed recommendation · Amazon Women also-viewed | error rate: 7.34% -> 7.37% | Hurt |
| Use PCD rather than Monomer | Amazon co-purchase recommendation · Amazon co-purchase | AUC: 0.8747 -> 0.8762 | Helped |
| Use PCD rather than Monomer | Polyvore recommendation · Polyvore average | AUC: 0.7405 -> 0.7443 | Helped |
| Increase the number of prototypes K while keeping total embedding dimension fixed (ablation) | PCD · Recommendation experiments | performance | Helped |
| Increase the total embedding size along with K (ablation) | PCD · Recommendation experiments | performance | No clear effect |
| Add the positive-pair distance regularizer controlled by lambda_m (ablation) | PCD recommendation · Recommendation experiments | recommendation performance | Hurt |
| Add the positive-pair distance regularizer controlled by lambda_m (ablation) | Compatibility Family generation · Generation experiments | generation quality | Helped |
| Increase the negative-pair margin m_prj (ablation) | MrCGAN · Generation experiments | output diversity | Hurt |
| Decrease the negative-pair margin m_prj (ablation) | MrCGAN · Generation experiments | incompatible-item generation | Hurt |
| Remove the positive generation margin m_enc (ablation) | MrCGAN · MNIST+1+2 | output diversity | Hurt |
Closest papers by meaning
- Deciphering Compatibility Relationships with Textual Descriptions via Extraction and Explanation (AAAI 2024 · similarity 0.87)
- ICAR: Image-Based Complementary Auto Reasoning (AAAI 2024 · similarity 0.84)
- Spectral-Based Graph Neural Networks for Complementary Item Recommendation (AAAI 2024 · similarity 0.82)
- Stationary Representations: Optimally Approximating Compatibility and Implications for Improved Model Replacements (CVPR 2024 · similarity 0.82)
- Optimize Incompatible Parameters Through Compatibility-aware Knowledge Integration (AAAI 2025 · similarity 0.81)
- Rethinking Semantic ID Construction for Generative Recommendation: SimHash with Parallel Decoding and Semantic Alignment (NeurIPS 2026 · similarity 0.81)
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.