Modeling User Preferences as Distributions for Optimal Transport-based Cross-domain Recommendation under Non-overlapping Settings
Ziyin Xiao, Toyotaro Suzumura
TL;DR
<3-5 sentence high-level summary> This paper tackles cross-domain recommendation without overlapping users or items by modeling user preferences as Gaussian Mixture Models and aligning domains through Optimal Transport. The proposed DUP-OT framework integrates a shared preprocessing stage, domain-specific GMM weight learning, and OT-based cross-domain alignment to transfer preferences at inference. Experiments on Amazon Review datasets demonstrate that distribution-based user representations plus OT alignment improve target-domain rating predictions and outperform non-overlapping baselines. The approach offers a scalable, privacy-friendly solution for cross-domain knowledge transfer in realistic settings.
Abstract
Cross-domain recommender (CDR) systems aim to transfer knowledge from data-rich domains to data-sparse ones, alleviating sparsity and cold-start issues present in conventional single-domain recommenders. However, many CDR approaches rely on overlapping users or items to establish explicit cross-domain connections, which is unrealistic in practice. Moreover, most methods represent user preferences as fixed discrete vectors, limiting their ability to capture the fine-grained and multi-aspect nature of user interests. To address these limitations, we propose DUP-OT (Distributional User Preferences with Optimal Transport), a novel framework for non-overlapping CDR. DUP-OT consists of three stages: (1) a shared preprocessing module that extracts review-based embeddings using a unified sentence encoder and autoencoder; (2) a user preference modeling module that represents each user's interests as a Gaussian Mixture Model (GMM) over item embeddings; and (3) an optimal-transport-based alignment module that matches Gaussian components across domains, enabling effective preference transfer for target-domain rating prediction. Experiments on Amazon Review datasets demonstrate that DUP-OT mitigates domain discrepancy and significantly outperforms state-of-the-art baselines under strictly non-overlapping training settings, with user correspondence revealed only for inference-time evaluation.