Generalized coupled symmetric tensor factorization for link prediction

Beyza Ermis, Ali Taylan Cemgil, Evrim Acar

Abstract: This study deals with the missing link prediction, the problem of predicting the existence of missing connections between entities of interest. Link prediction is addressed using coupled analysis of relational datasets represented by several matrices, including symmetric ones and multiway arrays, that will be simply called tensors. We propose to use an approach based on probabilistic interpretation of tensor factorisation models, i.e., Generalised Coupled Tensor Factorisation (GCTF), which can simultaneously fit a large class of tensor models to higher-order tensors/matrices with common latent factors using different loss functions. In addition, we propose the algorithm for factorization of symmetric matrices. Numerical experiments demonstrate that joint analysis of data from multiple sources via coupled factorisation and integration of symmetric matrices to models improves the link prediction performance and the selection of right loss function and tensor model is crucial for accurately predicting missing links.