Research map / Architectures and efficiency / Attention and transformers
Dual Sparse Attention Network For Session-based Recommendation
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 |
|---|---|---|---|
| add a learned target embedding as attention query rather than using the last-click (ablation) | DSAN · Diginetica | HR@20: 0.5199 -> 0.5376 | Helped |
| add a learned target embedding as attention query rather than using the last-click (ablation) | DSAN · Retailrocket | HR@20: 0.5322 -> 0.5654 | Helped |
| add a target attention layer to form the session representation (ablation) | DSAN · Diginetica | HR@20: 0.5348 -> 0.5376 | Helped |
| add a target attention layer to form the session representation (ablation) | DSAN · Retailrocket | HR@20: 0.5646 -> 0.5654 | Helped |
| combine self-attention and target attention sequentially, using learned target embedding instead of last-click as query (ablation) | DSAN · Diginetica | HR@20: 0.5340 -> 0.5376 | Helped |
| combine self-attention and target attention sequentially, using learned target embedding instead of last-click as query (ablation) | DSAN · Retailrocket | HR@20: 0.5600 -> 0.5654 | Helped |
| use adaptive α-entmax instead of softmax for attention weights (ablation) | DSAN · Diginetica and Retailrocket | HR@20 | Helped |
| learn α adaptively from session context rather than fixing α (ablation) | DSAN · Diginetica and Retailrocket | HR@20 | Helped |
| increase normalized weight wk from 1 to 20 (ablation) | DSAN · Diginetica and Retailrocket | HR@20 | Helped |
| increase normalized weight wk beyond 20 (ablation) | DSAN · Diginetica and Retailrocket | HR@20 | Hurt |
Closest papers by meaning
- LS-TGNN: Long and Short-Term Temporal Graph Neural Network for Session-Based Recommendation (AAAI 2025 · similarity 0.88)
- ScoreNet: Consistency-driven Framework with Multi-side Information Fusion for Session-based Recommendation (AAAI 2025 · similarity 0.87)
- Dynamic Multi-Interest Graph Neural Network for Session-Based Recommendation (AAAI 2025 · similarity 0.87)
- Iterative Sparse Attention for Long-sequence Recommendation (AAAI 2025 · similarity 0.87)
- Long-Sequence Recommendation Models Need Decoupled Embeddings (ICLR 2025 · similarity 0.87)
- Top-one Recommendation with Anonymous User Behaviors (Student Abstract) (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.