Contrastive Sequential Interaction Network Learning on Co-Evolving Riemannian Spaces

TL;DR

The paper tackles sequential interaction prediction in bipartite graphs by arguing that user and item representations should live in distinct, time-evolving Riemannian spaces rather than a single fixed space. It introduces CSincere, a framework with a Co-Evolving GNN and CurvNN curvature estimator, enabling cross-space message passing and curvature-driven evolution, paired with a Reweighed Co-Contrastive learning objective that leverages temporal views without labels. Empirical results on five public datasets show substantial gains over state-of-the-art baselines, validating the benefits of dual-curvature dynamics and cross-space interactions for SINs. The work advances curvature-aware, self-supervised representation learning for dynamic bipartite graphs, with practical implications for recommender systems and related temporal networks.

Abstract

The sequential interaction network usually find itself in a variety of applications, e.g., recommender system. Herein, inferring future interaction is of fundamental importance, and previous efforts are mainly focused on the dynamics in the classic zero-curvature Euclidean space. Despite the promising results achieved by previous methods, a range of significant issues still largely remains open: On the bipartite nature, is it appropriate to place user and item nodes in one identical space regardless of their inherent difference? On the network dynamics, instead of a fixed curvature space, will the representation spaces evolve when new interactions arrive continuously? On the learning paradigm, can we get rid of the label information costly to acquire? To address the aforementioned issues, we propose a novel Contrastive model for Sequential Interaction Network learning on Co-Evolving RiEmannian spaces, CSINCERE. To the best of our knowledge, we are the first to introduce a couple of co-evolving representation spaces, rather than a single or static space, and propose a co-contrastive learning for the sequential interaction network. In CSINCERE, we formulate a Cross-Space Aggregation for message-passing across representation spaces of different Riemannian geometries, and design a Neural Curvature Estimator based on Ricci curvatures for modeling the space evolvement over time. Thereafter, we present a Reweighed Co-Contrast between the temporal views of the sequential network, so that the couple of Riemannian spaces interact with each other for the interaction prediction without labels. Empirical results on 5 public datasets show the superiority of CSINCERE over the state-of-the-art methods.

Figures (5)

• Figure 2: Illustration of Co-Evolving GNN. In practice, a sequence of interactions is divided into several batches according to the associated timestamps (denoted by different colors), and each batch is regarded as a time interval. Cross-Space Aggregation, Interaction Integration and the neural Curvature Estimator are elaborated in Sec. 3.1, Sec. 3.2 and Sec. 3.3, respectively. Accordingly, user/item embeddings and curvature are learned via the proposed contrastive learning approach and curvature optimization, respectively.
• Figure 3: Illustration of Cross-Space Aggregation. We give an example of utilizing Eq. (\ref{['user_aggregation']}) to update user embedding once.
• Figure 4: A toy example of the co-contrast strategy.
• Figure 5: Effect of embedding dimensions
• Figure 6: Effect of sample ratio on MOOC dataset.

Theorems & Definitions (1)

• proof