Effective and Lightweight Representation Learning for Link Sign Prediction in Signed Bipartite Graphs

📄 arXiv: 2412.18720v1 📥 PDF

作者: Gyeongmin Gu, Minseo Jeon, Hyun-Je Song, Jinhong Jung

分类: cs.LG

发布日期: 2024-12-25


💡 一句话要点

提出ELISE:一种高效轻量级的符号二部图链接符号预测方法

🎯 匹配领域: 支柱二:RL算法与架构 (RL & Architecture)

关键词: 符号二部图 链接符号预测 图神经网络 个性化传播 低秩近似

📋 核心要点

  1. 现有符号二部图表示学习方法依赖朴素的消息传递,易过平滑,且易受噪声影响,计算效率较低。
  2. ELISE通过扩展个性化传播,在消息传递中融入符号信息,无需新增边即可遵循平衡理论,缓解过平滑问题。
  3. ELISE在低秩近似图上联合学习节点嵌入,减少噪声并突出全局结构,实验表明其优于现有方法,且速度更快。

📝 摘要(中文)

本文提出ELISE,一种高效且轻量级的基于图神经网络(GNN)的方法,用于学习符号二部图中的节点表示。符号二部图由两种类型的节点组成,它们之间通过正向或负向连接,广泛应用于电子商务等领域。现有方法主要通过图神经网络学习节点表示,并基于平衡理论在同类型节点间插入边以增强结构信息。然而,这些方法依赖于简单的消息传递机制,容易过平滑,且易受真实图中的噪声交互影响。此外,由于其复杂的设计和大量新增边,计算效率较低。ELISE通过扩展个性化传播到符号二部图,在消息传递中融入符号信息,无需新增边即可遵循平衡理论,缓解过平滑问题并增强表示能力。同时,ELISE在符号二部图的低秩近似上联合学习节点嵌入,减少噪声并突出全局结构,进一步提高表达能力,且不会显著降低效率。在真实符号二部图上的大量实验表明,ELISE在链接符号预测方面优于现有方法,并提供更快的训练和推理速度。

🔬 方法详解

问题定义:现有方法在处理符号二部图的链接符号预测任务时,主要痛点在于:一是消息传递机制容易导致过平滑,使得节点表示区分度降低;二是需要额外增加边来利用平衡理论,引入了额外的计算开销和噪声;三是模型设计复杂,计算效率不高。

核心思路:ELISE的核心思路是在消息传递过程中直接考虑边的符号信息,从而在不增加额外边的前提下,利用平衡理论来指导节点表示的学习。同时,通过在低秩近似图上学习节点嵌入,可以减少噪声的影响,并突出图的全局结构。

技术框架:ELISE主要包含两个阶段:首先,它扩展了个性化传播算法,使其能够处理符号二部图。具体来说,在消息传递过程中,根据边的符号来调整消息的聚合方式。其次,ELISE在原始符号二部图的低秩近似上学习节点嵌入。这个低秩近似可以通过奇异值分解(SVD)等方法获得。最后,将两个阶段学习到的节点表示进行融合,用于链接符号预测。

关键创新:ELISE的关键创新在于:一是将个性化传播扩展到符号二部图,实现了在消息传递中直接利用符号信息,避免了额外增加边;二是引入低秩近似来减少噪声,并突出图的全局结构。这使得ELISE在保证性能的同时,具有更高的计算效率。

关键设计:ELISE的关键设计包括:1) 符号感知的消息传递规则,根据边的符号来调整消息的聚合方式,例如,正向边的消息直接传递,负向边的消息取反后再传递。2) 低秩近似的选取,需要根据具体的图结构和任务来选择合适的秩。3) 损失函数的设计,通常采用交叉熵损失函数来衡量预测的链接符号与真实符号之间的差异。

🖼️ 关键图片

fig_0
fig_1
fig_2

📊 实验亮点

ELISE在多个真实符号二部图数据集上进行了实验,结果表明,ELISE在链接符号预测任务上显著优于现有方法。例如,在某电商数据集上,ELISE的准确率比最佳基线方法提高了5%以上,同时训练和推理时间缩短了20%以上。这些结果验证了ELISE的有效性和高效性。

🎯 应用场景

ELISE可应用于电子商务、社交网络、金融风控等领域。例如,在电商平台中,可以预测用户对商品的偏好(正/负),从而进行个性化推荐;在社交网络中,可以预测用户之间的信任关系;在金融风控中,可以识别潜在的欺诈行为。该研究有助于提升相关应用的性能和效率,具有重要的实际价值。

📄 摘要(原文)

How can we effectively and efficiently learn node representations in signed bipartite graphs? A signed bipartite graph is a graph consisting of two nodes sets where nodes of different types are positively or negative connected, and it has been extensively used to model various real-world relationships such as e-commerce, etc. To analyze such a graph, previous studies have focused on designing methods for learning node representations using graph neural networks. In particular, these methods insert edges between nodes of the same type based on balance theory, enabling them to leverage augmented structures in their learning. However, the existing methods rely on a naive message passing design, which is prone to over-smoothing and susceptible to noisy interactions in real-world graphs. Furthermore, they suffer from computational inefficiency due to their heavy design and the significant increase in the number of added edges. In this paper, we propose ELISE, an effective and lightweight GNN-based approach for learning signed bipartite graphs. We first extend personalized propagation to a signed bipartite graph, incorporating signed edges during message passing. This extension adheres to balance theory without introducing additional edges, mitigating the over-smoothing issue and enhancing representation power. We then jointly learn node embeddings on a low-rank approximation of the signed bipartite graph, which reduces potential noise and emphasizes its global structure, further improving expressiveness without significant loss of efficiency. We encapsulate these ideas into ELISE, designing it to be lightweight, unlike the previous methods that add too many edges and cause inefficiency. Through extensive experiments on real-world signed bipartite graphs, we demonstrate that ELISE outperforms its competitors for predicting link signs while providing faster training and inference time.