学术讲座

大数据管理与分析方法研究北京市重点实验室（BDAI）研究生沙龙由中国人民大学高瓴人工智能学院师生组织定期举行。4月13日研讨会由赵鑫教授指导的学生侯宇蓬以及高瓴人工智能学院博士后张骁介绍自己的研究工作。欢迎同学们积极参与研讨！

报告题目: Improving General and Session-based Recommendation via Contrastive Learning

报告时间：4月13日12:30-13:00

报告地点：立德楼801

个人简介: 侯宇蓬，研究生二年级

导师：赵鑫 教授

研究方向：推荐系统与图神经网络

报告摘要: This talk will present our recent research about how to improve graph collaborative filtering and session-based recommendation via contrastive learning. Firstly, though graph collaborative filtering methods have been proposed as an effective recommendation approach, these methods suffer from data sparsity in real scenarios. We propose NCL, a neighborhood-enriched contrastive learning approach, which explicitly incorporates the potential structural and semantic neighbors into contrastive pairs. NCL is a model-agnostic plugin method, notably with 26% and 17% performance gain over a competitive graph collaborative filtering base model on the Yelp and Amazon-book datasets, respectively. Secondly, we observed that session embedding encoded by non-linear encoder is usually not in the same representation space as item embeddings, resulting in the inconsistent prediction issue. We propose CORE to unify the representation space throughout the encoding and decoding process in session-based recommendation. Additionally, I'll show how the open-source recommendation library RecBole can help with conducting these two works.