Multi-way Clustering and Discordance Analysis through Deep Collective Matrix Tri-Factorization
This addresses the challenge of exploratory analysis for heterogeneous multimodal relational data in domains like knowledge bases, though it appears incremental as it builds on existing collective matrix factorization methods with neural enhancements.
The paper tackles the problem of analyzing heterogeneous multi-typed relational data by developing a neural method for collective matrix tri-factorization to perform spectral clustering of all entities and learn cluster associations, achieving improved performance over non-neural approaches on benchmark datasets. It also introduces Discordance Analysis to reveal information discrepancies across matrix subsets, demonstrating utility in knowledge base quality assessment and representation learning.
Heterogeneous multi-typed, multimodal relational data is increasingly available in many domains and their exploratory analysis poses several challenges. We advance the state-of-the-art in neural unsupervised learning to analyze such data. We design the first neural method for collective matrix tri-factorization of arbitrary collections of matrices to perform spectral clustering of all constituent entities and learn cluster associations. Experiments on benchmark datasets demonstrate its efficacy over previous non-neural approaches. Leveraging signals from multi-way clustering and collective matrix completion we design a unique technique, called Discordance Analysis, to reveal information discrepancies across subsets of matrices in a collection with respect to two entities. We illustrate its utility in quality assessment of knowledge bases and in improving representation learning.