Transfer Learning Under High-Dimensional Network Convolutional Regression Model

2025-04-28

AI Summary

Transfer learning for high-dimensional network data faces dual challenges: severe label scarcity in the target domain and structural dependencies among nodes. Method: This paper proposes Network Convolutional Regression (NCR), a transfer framework that models local dependencies between node responses and their own as well as neighbor features. It introduces a two-stage transfer algorithm and a source-domain adaptive selection mechanism. Under the Erdős–Rényi random graph assumption, we provide the first theoretical proof that transfer accelerates Lasso estimation convergence. Contribution/Results: By integrating cross-network domain shift correction and structure-aware regularization, NCR significantly improves prediction accuracy on both synthetic benchmarks and real-world Sina Weibo data. Its gains are especially pronounced when the target domain contains only a handful of labeled nodes. The framework thus offers both theoretical rigor—establishing convergence-rate benefits of transfer—and practical efficacy—demonstrated via empirical superiority over state-of-the-art baselines.

Technology Category

• Machine Learning: Transfer, Domain Adaptation, Multi-Task Learning
• Application Domains: Humanities & Computational Social Science
• 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
• Social Networks and Social Media: Influence propagation, information diffusion, and the prediction on networks
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

Transfer learning enhances model performance by utilizing knowledge from related domains, particularly when labeled data is scarce. While existing research addresses transfer learning under various distribution shifts in independent settings, handling dependencies in networked data remains challenging. To address this challenge, we propose a high-dimensional transfer learning framework based on network convolutional regression (NCR), inspired by the success of graph convolutional networks (GCNs). The NCR model incorporates random network structure by allowing each node's response to depend on its features and the aggregated features of its neighbors, capturing local dependencies effectively. Our methodology includes a two-step transfer learning algorithm that addresses domain shift between source and target networks, along with a source detection mechanism to identify informative domains. Theoretically, we analyze the lasso estimator in the context of a random graph based on the Erdos-Renyi model assumption, demonstrating that transfer learning improves convergence rates when informative sources are present. Empirical evaluations, including simulations and a real-world application using Sina Weibo data, demonstrate substantial improvements in prediction accuracy, particularly when labeled data in the target domain is limited.

Problem

• Transfer learning for high-dimensional networked data dependencies
• Addressing domain shift in source and target network data
• Improving prediction accuracy with limited labeled target data

Innovation

• High-dimensional transfer learning with network convolutional regression
• Two-step algorithm addressing domain shift and source detection
• Lasso estimator analysis under Erdos-Renyi graph model