GRVFL-MV: Graph Random Vector Functional Link Based on Multi-View Learning

2024-09-07

Information Sciences

AI Summary

Random Vector Functional Link (RVFL) networks, constrained by their shallow architecture, struggle to capture geometric properties of graph-structured data and inter-view correlations in multi-view settings. Method: We propose the Graph Random Vector Functional Link Multi-View (GRVFL-MV) model—the first RVFL extension to multi-view graph learning. GRVFL-MV jointly incorporates graph topological priors and multi-view features within a single-step analytical solution framework, integrating cross-view consistency regularization and graph Laplacian constraints—bypassing iterative gradient-based optimization. Results: Evaluated on six standard multi-view graph benchmarks, GRVFL-MV achieves average classification accuracy gains of 3.2–7.8% over state-of-the-art baselines. It trains 12–28× faster than GNN-based methods, reduces parameter count by over 99%, and simultaneously enhances computational efficiency, generalization capability, and model interpretability.

Technology Category

• Machine Learning: Multi-instance/Multi-view Learning
• Computer Vision: Multi-modal Vision
• Natural Language Processing: Language Grounding & Multi-modal NLP

Application Category

• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Problem

• Enhance RVFL classification performance
• Incorporate multi-view and geometrical data properties
• Improve generalization across diverse datasets

Innovation

• Graph embedding enhances data structure
• Multi-view learning captures complex patterns
• RVFL networks improve nonlinear relationship detection