Adjusted Count Quantification Learning on Graphs

2025-03-12

AI Summary

Existing node clustering methods for graph-structured data fail to satisfy the prior probability shift assumption required by Adjusted Class Counting (ACC), limiting their applicability to label distribution prediction in graph quantification. Method: This work extends the ACC framework to graph quantification for the first time. It introduces Structural Importance Sampling (SIS) to mitigate covariate shift and designs a neighborhood-aware ACC method to suppress interference from non-homophilous edges. Furthermore, it establishes the first theoretical correction model for prior probability shift in graph quantification. Results: The proposed approach significantly outperforms state-of-the-art methods across multiple graph quantification benchmarks, empirically validating the robustness and generalizability of SIS and neighborhood-aware ACC on real-world graph data.

Technology Category

• Machine Learning: Graph-based Machine Learning
• Reasoning under Uncertainty: Graphical Models
• Data Mining & Knowledge Management: Graph Mining, Social Network Analysis & Community

Application Category

• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• Semantics and Knowledge: Methods, algorithms and applications for the development of semantic models, knowledge graphs and other forms of structured data models with machine-interpretable semantics
• Social Networks and Social Media: Influence propagation, information diffusion, and the prediction on networks

Abstract

Quantification learning is the task of predicting the label distribution of a set of instances. We study this problem in the context of graph-structured data, where the instances are vertices. Previously, this problem has only been addressed via node clustering methods. In this paper, we extend the popular Adjusted Classify&Count (ACC) method to graphs. We show that the prior probability shift assumption upon which ACC relies is often not fulfilled and propose two novel graph quantification techniques: Structural importance sampling (SIS) makes ACC applicable in graph domains with covariate shift. Neighborhood-aware ACC improves quantification in the presence of non-homophilic edges. We show the effectiveness of our techniques on multiple graph quantification tasks.

Problem

• Predict label distribution of graph vertices.
• Extend Adjusted Classify & Count to graphs.
• Address covariate shift and non-homophilic edges.

Innovation

• Extends Adjusted Classify & Count to graphs
• Introduces Structural Importance Sampling for covariate shift
• Develops Neighborhood-aware ACC for non-homophilic edges