Bayesian multi-tensor factorization
This work addresses the need for unsupervised multi-view learning from multiple data tensors in fields like bioinformatics and neuroscience, representing an incremental advancement by extending existing factorization methods to a Bayesian framework.
The authors tackled the problem of jointly factorizing multiple matrices and tensors by introducing the first Bayesian formulation for this task, which generalizes to arbitrary sets of tensors and relaxes trilinear assumptions, resulting in a factorization into shared and private factors. They demonstrated performance against existing baselines in structural toxicogenomics and functional neuroimaging tasks.
We introduce Bayesian multi-tensor factorization, a model that is the first Bayesian formulation for joint factorization of multiple matrices and tensors. The research problem generalizes the joint matrix-tensor factorization problem to arbitrary sets of tensors of any depth, including matrices, can be interpreted as unsupervised multi-view learning from multiple data tensors, and can be generalized to relax the usual trilinear tensor factorization assumptions. The result is a factorization of the set of tensors into factors shared by any subsets of the tensors, and factors private to individual tensors. We demonstrate the performance against existing baselines in multiple tensor factorization tasks in structural toxicogenomics and functional neuroimaging.