Core Knowledge Learning Framework for Graph Adaptation and Scalability Learning

TL;DR

The paper introduces Core Knowledge Learning (CKL), a framework that learns a task-relevant core subgraph to address domain shift and data scarcity in graph classification. By extracting G_sub via node/edge selection and using it for graph domain adaptation and few-shot learning, CKL achieves improved robustness and scalability compared to state-of-the-art methods. The approach integrates mutual-information-based explainability, WL-subtree kernel-based domain transfer, and bi-level optimization for few-shot tasks, demonstrating strong empirical gains across diverse graph datasets and molecular tasks. CKL further shows flexibility with different GNN backbones and kernels, highlighting its potential as a unified solution for cross-domain graph learning problems.

Abstract

Graph classification is a pivotal challenge in machine learning, especially within the realm of graph-based data, given its importance in numerous real-world applications such as social network analysis, recommendation systems, and bioinformatics. Despite its significance, graph classification faces several hurdles, including adapting to diverse prediction tasks, training across multiple target domains, and handling small-sample prediction scenarios. Current methods often tackle these challenges individually, leading to fragmented solutions that lack a holistic approach to the overarching problem. In this paper, we propose an algorithm aimed at addressing the aforementioned challenges. By incorporating insights from various types of tasks, our method aims to enhance adaptability, scalability, and generalizability in graph classification. Motivated by the recognition that the underlying subgraph plays a crucial role in GNN prediction, while the remainder is task-irrelevant, we introduce the Core Knowledge Learning (\method{}) framework for graph adaptation and scalability learning. \method{} comprises several key modules, including the core subgraph knowledge submodule, graph domain adaptation module, and few-shot learning module for downstream tasks. Each module is tailored to tackle specific challenges in graph classification, such as domain shift, label inconsistencies, and data scarcity. By learning the core subgraph of the entire graph, we focus on the most pertinent features for task relevance. Consequently, our method offers benefits such as improved model performance, increased domain adaptability, and enhanced robustness to domain variations. Experimental results demonstrate significant performance enhancements achieved by our method compared to state-of-the-art approaches.

Figures (5)

• Figure 1: Illustration of the core knowledge learning framework. The framework extracts the core subgraph from the entire graph, which represents the fundamental structure necessary for task-relevant predictions. This core subgraph is then utilized for downstream tasks, including graph domain adaptation and few-shot learning tasks.
• Figure 2: An overview of the proposed CKL. CKL first utilizes the node embeddings for node selection, and then cooperates the edge embeddings for edge selection to obtain the core subgraph.
• Figure 3: The core subgraph in the graph domain adaptation task. CKL employs a kernel function to assess the similarity between source and target subgraphs, and assigning labels to the target graphs based on the most similar source graph.
• Figure 4: The core subgraph in the few-shot learning task. CKL employs the bi-level method to optimize the core subgraph learning and multi-task prediction.
• Figure 5: The performance with different GNNs and kernels on different datasets. (a), (b) are the performance of different GNNs, (c), (d) are the performance of different graph kernels.