Multi-view biclustering via non-negative matrix tri-factorisation

2025-02-19

AI Summary

To address the heterogeneity in feature–sample compatibility across multiple views, this paper proposes ResNMTF—a novel multi-view biclustering framework that automatically determines the number of clusters. ResNMTF jointly identifies sample subsets and their view-specific feature subsets, while supporting arbitrary shared-dimension configurations across views. Methodologically, it integrates nonnegative matrix tri-factorization with multi-view co-clustering and introduces the bisilhouette score—a first-of-its-kind evaluation metric specifically designed for biclustering structures—enabling both hyperparameter self-selection and result visualization. Extensive experiments on synthetic and real-world datasets demonstrate that ResNMTF accurately recovers overlapping and incomplete biclusters. The bisilhouette score exhibits strong agreement with external evaluation metrics (average correlation > 0.92), confirming its validity and practical utility for multi-view biclustering.

Technology Category

• Machine Learning: Multi-instance/Multi-view Learning
• Computer Vision: Multi-modal Vision
• Natural Language Processing: Machine Translation, Multilinguality, Cross-Lingual NLP

Application Category

• Graph Algorithms and Modeling for the Web: Representation, reconstruction, and subgraph or motif discovery in Web-related graphs
• Web Mining and Content Analysis: Normalization, clustering, classification, and summarization of Web text
• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics

Abstract

Multi-view data is ever more apparent as methods for production, collection and storage of data become more feasible both practically and fiscally. However, not all features are relevant to describe the patterns for all individuals. Multi-view biclustering aims to simultaneously cluster both rows and columns, discovering clusters of rows as well as their view-specific identifying features. A novel multi-view biclustering approach based on non-negative matrix factorisation is proposed (ResNMTF). Demonstrated through extensive experiments on both synthetic and real datasets, ResNMTF successfully identifies both overlapping and non-exhaustive biclusters, without pre-existing knowledge of the number of biclusters present, and is able to incorporate any combination of shared dimensions across views. Further, to address the lack of a suitable bicluster-specific intrinsic measure, the popular silhouette score is extended to the bisilhouette score. The bisilhouette score is demonstrated to align well with known extrinsic measures, and proves useful as a tool for hyperparameter tuning as well as visualisation.

Problem

• Identifies overlapping and non-exhaustive biclusters without prior knowledge
• Clusters rows and view-specific features simultaneously in multi-view data
• Extends silhouette score to bisilhouette for bicluster evaluation and tuning

Innovation

• Non-negative matrix tri-factorisation for multi-view biclustering
• ResNMTF identifies overlapping non-exhaustive biclusters
• Extended silhouette score to bisilhouette for evaluation