AdaMCT: Adaptive Mixture of CNN-Transformer for Sequential Recommendation

By Juyong Jiang et al

Read the original document by opening this link in a new tab.

Table of Contents

1. Introduction
2. Related Work
3. Methodology
3.1 Problem Statement
3.2 Model Architecture
3.3 Embedding Module
3.4 Adaptive Mixture of CNN-Transformer
3.4.1 Global Attention Module
3.4.2 Local Convolutional Module
4. Results
5. Conclusion

Summary

Sequential recommendation (SR) aims to model users’ dynamic preferences from a series of interactions. The AdaMCT model combines CNN and Transformer mechanisms to capture both long-term and short-term preferences effectively. Adaptive mixture units enhance expressibility and personalized user modeling. Experimental results demonstrate the effectiveness of AdaMCT in comparison to existing models.