Xiuyuan Qin

Xiuyuan Qin

Sequential recommendation & contrastive learning

Meta-optimized Contrastive Learning for Sequential Recommendation

Published in SIGIR 2023 (Oral), 2023

Recommended citation: Xiuyuan Qin, Huanhuan Yuan, Pengpeng Zhao, Junhua Fang, Fuzhen Zhuang, Guanfeng Liu, Yanchi Liu, and Victor S. Sheng. 2023. Meta-optimized Contrastive Learning for Sequential Recommendation. In SIGIR. 89–98. https://arxiv.org/abs/2304.07763

Venue: The 46th International ACM SIGIR Conference on Research and Development in Information Retrieval (SIGIR 2023), Oral presentation.

Resources: Paper (arXiv) · Code · Video · Note

Abstract

Contrastive Learning (CL) is a rising approach for sparse and noisy recommendation data. Most existing CL methods rely on hand-crafted data or model augmentation and often need large batches or memory banks. MCLRec applies both data augmentation and learnable model augmentation, contrasting data and model augmented views, and uses meta-learning to guide augmenter updates. A contrastive regularization encourages informative views and avoids overly similar pairs. Experiments on commonly used datasets validate the effectiveness of MCLRec.

Citation

Xiuyuan Qin, Huanhuan Yuan, Pengpeng Zhao, Junhua Fang, Fuzhen Zhuang, Guanfeng Liu, Yanchi Liu, and Victor S. Sheng. 2023. Meta-optimized Contrastive Learning for Sequential Recommendation. In SIGIR ’23. 89–98. DOI · arXiv:2304.07763.