Summary of Spatial-temporal Knowledge Distillation For Takeaway Recommendation, by Shuyuan Zhao et al.
Spatial-Temporal Knowledge Distillation for Takeaway Recommendation
by Shuyuan Zhao, Wei Chen, Boyan Shi, Liyong Zhou, Shuohao Lin, Huaiyu Wan
First submitted to arxiv on: 21 Dec 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Information Retrieval (cs.IR)
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary The proposed spatial-temporal knowledge distillation model for takeaway recommendation (STKDRec) aims to improve user satisfaction by recommending future purchases based on historical behaviors, while also boosting merchant sales. The existing methods focus on incorporating auxiliary information or leveraging knowledge graphs to alleviate the sparsity issue of user purchase sequences. However, these approaches are limited by two main challenges: capturing dynamic user preferences on complex geospatial information and efficiently integrating spatial-temporal knowledge from both graphs and sequence data with low computational costs. To address this, the authors propose a novel two-stage training process comprising a spatial-temporal knowledge graph (STKG) encoder and a spatial-temporal Transformer (ST-Transformer). The STKD strategy is introduced to transfer graph-based spatial-temporal knowledge to the ST-Transformer, facilitating the adaptive fusion of rich knowledge derived from both the STKG and sequence data while reducing computational overhead. Experimental results on three real-world datasets show that STKDRec significantly outperforms state-of-the-art baselines. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Takeaway recommendation systems aim to suggest users’ future purchases based on their past behaviors, making it easier for customers to find what they like and boosting sales for merchants. Existing methods try to improve these recommendations by using extra information or special types of graphs. However, there are two main challenges: understanding how users’ preferences change over time and space, and combining this information with user purchase patterns in a way that’s efficient and accurate. The authors propose a new approach that uses both spatial-temporal knowledge graphs and sequence data to make better recommendations. They show that their method outperforms current approaches on three real-world datasets. |
Keywords
» Artificial intelligence » Boosting » Encoder » Knowledge distillation » Knowledge graph » Transformer
