Deep collective matrix factorization for augmented multi-view learning

DOI10.1007/S10994-019-05801-6MaRDI QIDQ2320570zbMATH OpenOpenAlexWikidataFDO

Authors Ragunathan Mariappan, Vaibhav Rajan

Publication date 23 August 2019

Published in Machine Learning (Search for Journal in Brave)

Full work available at URL https://arxiv.org/abs/1811.11427

zbMATH Keywords

augmented multi-view learningBayesian optimizationcollective matrix factorizationdeep learninggene-disease prioritizationrecommendation

Mathematics Subject Classification ID

Applications of statistics to biology and medical sciences; meta analysis (62P10) Learning and adaptive systems in artificial intelligence (68T05) Artificial neural networks and deep learning (68T07)

Abstract: Learning by integrating multiple heterogeneous data sources is a common requirement in many tasks. Collective Matrix Factorization (CMF) is a technique to learn shared latent representations from arbitrary collections of matrices. It can be used to simultaneously complete one or more matrices, for predicting the unknown entries. Classical CMF methods assume linearity in the interaction of latent factors which can be restrictive and fails to capture complex non-linear interactions. In this paper, we develop the first deep-learning based method, called dCMF, for unsupervised learning of multiple shared representations, that can model such non-linear interactions, from an arbitrary collection of matrices. We address optimization challenges that arise due to dependencies between shared representations through Multi-Task Bayesian Optimization and design an acquisition function adapted for collective learning of hyperparameters. Our experiments show that dCMF significantly outperforms previous CMF algorithms in integrating heterogeneous data for predictive modeling. Further, on two tasks - recommendation and prediction of gene-disease association - dCMF outperforms state-of-the-art matrix completion algorithms that can utilize auxiliary sources of information.

Review text

The existence of multiple heterogeneous information represented by an arbitrary collection of matrices is very common in modern data analysis. An exploitation of this heterogeneous information can boost the learning performance, and is an important machine learning task. The existing approaches assume a linear interaction of latent factors, and are therefore not applicable to problems with nonlinear interactions. In this paper, the authors propose a first deep-learning method for collective factorization of matrices. Experimental results demonstrate the advantage of the proposed method over existing methods in the application domains of recommendation and gene-disease association prediction.

Recommendations

• Multi-relational learning via hierarchical nonparametric Bayesian collective matrix factorization
• A survey on deep matrix factorizations
• Collective matrix completion
• Bayesian multi-tensor factorization
• Low-rank matrix factorization with multiple hypergraph regularizer

Cites work

• A taxonomy of global optimization methods based on response surfaces
• Canonical Correlation Analysis: An Overview with Application to Learning Methods
• Deep learning
• scientific article; zbMATH DE number 6276119 (Why is no real title available?)
• Learning Theory
• ParEGO
• Reducing the Dimensionality of Data with Neural Networks
• RELATIONS BETWEEN TWO SETS OF VARIATES
• Stacked denoising autoencoders: learning useful representations in a deep network with a local denoising criterion
• Statistical comparisons of classifiers over multiple data sets

Cited in

(4)

• Bayesian multi-tensor factorization
• A survey on deep matrix factorizations
• Multi-relational learning via hierarchical nonparametric Bayesian collective matrix factorization
• Multi-task learning for spatial events prediction from social data

Describes a project that uses

Uses Software

• ParEGO
• Hyperopt
• darch
• Spearmint

This page was built for publication: Deep collective matrix factorization for augmented multi-view learning

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q2320570)