---
tags: 暑假
---

# 0722
> https://airtable.com/shrH8vpKlF8U0stSz

## 主題方向
### 大方向
- Sequential Recommendation
- predict user rating to the items
- 不能直接把rating當training feature
- Next-Item Recommendation
- Next-Basket Recommendation
- CTR-prediction

### 照主題
- **RNN based**
- Modeling Personalized Item Frequency Information for Next-basket Recommendation
- **純Attention**
- Attentive Sequential Model of Latent Intent for Next Item Recommendation
- A Contextualized Temporal Attention Mechanism for Sequential Recommendation
- Rethinking Item Importance in Session-based Recommendation
- **GNN + Attention (目前主流)**
- TAGNN: Target Attentive Graph Neural Networks for Session-based Recommendation
- **HyperGraph**
- Next-item Recommendation with Sequential Hypergraphs
- **Translation-Based**
- Seq2seq Translation Model for Sequential Recommendation
- Adaptive Hierarchical Translation-based Sequential Recommendation
- **Add item KG**
- Incorporating User Micro-behaviors and Item Knowledge into Multi-task Learning for Session-based Recommendation
- **Cold-start**
- MAMO: Memory-Augmented Meta-Optimization for Cold-start Recommendation
- **Streaming problem**
- GAG: Global Attributed Graph Neural Network for Streaming Session-based Recommendation
- **Training tricks**
- Future Data Helps Training: Modeling Future Contexts for Session-based Recommendation

## Github paperlists
- https://github.com/OnYuKang/Recommendation-systems-paperlist/blob/master/README.md

---
### 2018 & 2019:
- **Social(Graphsage想法)**
- Session-based Social Recommendation via Dynamic Graph Attention Networks
- **CNN**
- A Simple Convolutional Generative Network for Next Item Recommendation

- 2020已有的主題
- Attention-Based Transactional Context Embedding for Next-Item Recommendation
-

---
Temporal類的做法
1. SOTA
2. 跑baseline
3. 提主要idea

- 同item在不同時間意義會不同
- 同user/ session在不同時刻的preference不同
- 同items組成的session, 會因順序不同而有不同意義
- sliding window/ attention => 抓不同區段對item的影響