Tensor Factorization for Multi-relational Learning

Maximilian Nickel, Volker Tresp

Abstract: Tensor factorization has emerged as a promising approach for solving relational learning tasks. Here we review recent results on a particular tensor factorization approach, i.e. Rescal, which has demonstrated state-of-the-art relational learning results, while scaling to knowledge bases with millions of entities and billions of known facts.