Training A Foundation Model to Represent Graphs as Vectors

2026-02-04

AI Summary

This work proposes a universal graph foundation model designed to encode arbitrary graphs into vector representations that preserve both structural and semantic information, thereby supporting graph-level tasks and enabling cross-domain generalization. The approach integrates a multi-graph feature alignment mechanism with a density-maximized mean alignment algorithm to enhance consistency of node embeddings across datasets. Discriminative graph representations are learned through graph neural networks combined with contrastive learning, while a novel pooling-free, multi-layer reference distribution module efficiently aggregates node-level information into graph-level representations. Theoretical analysis provides an upper bound on the generalization error. Extensive experiments demonstrate that the model significantly outperforms strong baselines on few-shot graph classification and clustering tasks, validating its superior representational capacity and generalization ability.

Technology Category

• Machine Learning: Graph-based Machine Learning
• Reasoning under Uncertainty: Graphical Models
• Computer Vision: Representation Learning for Vision

Application Category

• Graph Algorithms and Modeling for the Web: Graph embeddings and representation learning for Web-related graphs
• Semantics and Knowledge: Scalable techniques for the creation, curation, publication, maintenance, and consumption of large, Web-based, structured, reusable, knowledge graphs and ontologies
• Search and Retrieval-Augmented AI: Web query analysis, representation and understanding

Abstract

This paper aims to train a graph foundation model that is able to represent any graph as a vector preserving structural and semantic information useful for downstream graph-level tasks such as graph classification and graph clustering. To learn the features of graphs from diverse domains while maintaining strong generalization ability to new domains, we propose a multi-graph-based feature alignment method, which constructs weighted graphs using the attributes of all nodes in each dataset and then generates consistent node embeddings. To enhance the consistency of the features from different datasets, we propose a density maximization mean alignment algorithm with guaranteed convergence. The original graphs and generated node embeddings are fed into a graph neural network to achieve discriminative graph representations in contrastive learning. More importantly, to enhance the information preservation from node-level representations to the graph-level representation, we construct a multi-layer reference distribution module without using any pooling operation. We also provide a theoretical generalization bound to support the effectiveness of the proposed model. The experimental results of few-shot graph classification and graph clustering show that our model outperforms strong baselines.

Problem

• graph foundation model
• graph representation
• structural information
• semantic information
• generalization

Innovation

• graph foundation model
• feature alignment
• density maximization
• contrastive learning
• reference distribution module