Separated Contrastive Learning for Matching in Cross-domain Recommendation with Curriculum Scheduling

TL;DR

This paper addresses instability in cross-domain matching caused by jointly optimizing intra-domain and inter-domain contrastive tasks. It introduces SCCDR, a framework that separates intra-CL and inter-CL into two stages and incorporates a stop-gradient operation to stabilize knowledge transfer, along with a curriculum scheduler that gradually increases the difficulty of inter-domain negatives based on a Katz centrality-based measure. The approach uses dual GraphSAGE encoders with intra-CL BCELoss and inter-CL aligned-user and neighbor-similarity losses, coupled via multi-task optimization with weights $\lambda_{intra}$ and $\lambda_{inter}$. Extensive experiments on multiple Amazon-domain pairs and a proprietary Industrial dataset, plus online A/B testing, show SCCDR achieving state-of-the-art performance in offline HIT@N metrics and significant online gains, especially in cold-start and less-correlated domain transfers. The work highlights the importance of modality-aware contrastive learning and curriculum strategies for robust cross-domain knowledge transfer in recommender systems.

Abstract

Cross-domain recommendation (CDR) is a task that aims to improve the recommendation performance in a target domain by leveraging the information from source domains. Contrastive learning methods have been widely adopted among intra-domain (intra-CL) and inter-domain (inter-CL) users/items for their representation learning and knowledge transfer during the matching stage of CDR. However, we observe that directly employing contrastive learning on mixed-up intra-CL and inter-CL tasks ignores the difficulty of learning from inter-domain over learning from intra-domain, and thus could cause severe training instability. Therefore, this instability deteriorates the representation learning process and hurts the quality of generated embeddings. To this end, we propose a novel framework named SCCDR built up on a separated intra-CL and inter-CL paradigm and a stop-gradient operation to handle the drawback. Specifically, SCCDR comprises two specialized curriculum stages: intra-inter separation and inter-domain curriculum scheduling. The former stage explicitly uses two distinct contrastive views for the intra-CL task in the source and target domains, respectively. Meanwhile, the latter stage deliberately tackles the inter-CL tasks with a curriculum scheduling strategy that derives effective curricula by accounting for the difficulty of negative samples anchored by overlapping users. Empirical experiments on various open-source datasets and an offline proprietary industrial dataset extracted from a real-world recommender system, and an online A/B test verify that SCCDR achieves state-of-the-art performance over multiple baselines.

Figures (5)

• Figure 1: The stability comparison between intra-CL and inter-CL losses during the training of cross-domain recommendation task on Amazon Books-Videos dataset.
• Figure 2: The illustration of our proposed framework SCCDR and the example of the intra-CL and inter-CL losses we considered.
• Figure 3: Sensitivity analysis on the loss weights $\lambda_{\text{intra}}$ and $\lambda_{\text{inter}}$ on Elec-Cloth dataset regarding both HIT@100 and AUC metrics.
• Figure 4: Books-Videos
• Figure 5: Elec-Cloth