ZeroG: Investigating Cross-dataset Zero-shot Transferability in Graphs

📄 arXiv: 2402.11235v2 📥 PDF

作者: Yuhan Li, Peisong Wang, Zhixun Li, Jeffrey Xu Yu, Jia Li

分类: cs.LG

发布日期: 2024-02-17 (更新: 2024-06-24)

备注: Accepted by SIGKDD 2024, research track

🔗 代码/项目: GITHUB


💡 一句话要点

提出ZeroG框架以解决图数据的跨数据集零-shot迁移问题

🎯 匹配领域: 支柱九:具身大模型 (Embodied Foundation Models)

关键词: 零-shot学习 图学习 迁移学习 语言模型 子图采样 特征对齐 模型泛化

📋 核心要点

  1. 现有图学习方法在面对新图数据时,往往需要特定的数据集和标签进行微调,限制了其通用性。
  2. ZeroG框架通过语言模型编码节点属性和类别语义,解决特征不对齐和标签空间不匹配的问题,增强了跨数据集的迁移能力。
  3. 实验结果显示,ZeroG在多个数据集上实现了显著的零-shot迁移效果,提升了模型的泛化能力。

📝 摘要(中文)

随着基础模型(如大型语言模型)的发展,零-shot迁移学习变得愈发重要。本文提出了ZeroG框架,旨在实现图学习中的跨数据集零-shot迁移。该框架通过语言模型编码节点属性和类别语义,解决特征不对齐、标签空间不匹配和负迁移等挑战。此外,提出的基于提示的子图采样模块增强了提取子图的语义和结构信息。实验结果表明,ZeroG在跨数据集零-shot迁移方面表现出显著的有效性,为图基础模型的发展开辟了新路径。

🔬 方法详解

问题定义:本文旨在解决图学习中跨数据集零-shot迁移的挑战。现有方法通常依赖于特定数据集的微调,导致在新图数据上表现不佳,面临特征不对齐和标签空间不匹配等问题。

核心思路:ZeroG框架通过利用语言模型来编码节点属性和类别语义,确保不同数据集之间特征维度的一致性,从而实现更好的迁移学习效果。

技术框架:ZeroG的整体架构包括三个主要模块:1) 语言模型编码模块,用于提取节点特征;2) 基于提示的子图采样模块,增强子图的语义和结构信息;3) 轻量级微调策略,降低过拟合风险。

关键创新:ZeroG的核心创新在于引入了基于提示的子图采样模块,利用提示节点和邻域聚合来丰富子图的语义信息,这一设计显著提升了模型的迁移能力。

关键设计:在模型设计中,采用了轻量级的微调策略,以保持语言模型的零-shot学习效能,同时在损失函数和网络结构上进行了优化,以适应不同数据集的特征。

🖼️ 关键图片

fig_0
fig_1
fig_2

📊 实验亮点

实验结果表明,ZeroG在多个数据集上实现了显著的零-shot迁移效果,相较于基线模型,迁移准确率提升了20%以上,验证了其在跨数据集学习中的有效性。

🎯 应用场景

ZeroG框架在图学习领域具有广泛的应用潜力,特别是在社交网络分析、生物信息学和推荐系统等领域。通过实现跨数据集的零-shot迁移,ZeroG能够降低对标注数据的依赖,提高模型在新场景下的适应能力,推动图基础模型的进一步发展。

📄 摘要(原文)

With the development of foundation models such as large language models, zero-shot transfer learning has become increasingly significant. This is highlighted by the generative capabilities of NLP models like GPT-4, and the retrieval-based approaches of CV models like CLIP, both of which effectively bridge the gap between seen and unseen data. In the realm of graph learning, the continuous emergence of new graphs and the challenges of human labeling also amplify the necessity for zero-shot transfer learning, driving the exploration of approaches that can generalize across diverse graph data without necessitating dataset-specific and label-specific fine-tuning. In this study, we extend such paradigms to zero-shot transferability in graphs by introducing ZeroG, a new framework tailored to enable cross-dataset generalization. Addressing the inherent challenges such as feature misalignment, mismatched label spaces, and negative transfer, we leverage a language model to encode both node attributes and class semantics, ensuring consistent feature dimensions across datasets. We also propose a prompt-based subgraph sampling module that enriches the semantic information and structure information of extracted subgraphs using prompting nodes and neighborhood aggregation, respectively. We further adopt a lightweight fine-tuning strategy that reduces the risk of overfitting and maintains the zero-shot learning efficacy of the language model. The results underscore the effectiveness of our model in achieving significant cross-dataset zero-shot transferability, opening pathways for the development of graph foundation models. Codes and data are available at https://github.com/NineAbyss/ZeroG.