Bilkent University
Department of Computer Engineering


An Efficient Model for Distributed Sparse Tensor Factorization


Tuğba Torun
PhD Student
Computer Engineering Department
Bilkent University

Tensors are the structures that represent multi-dimensional data which can be considered as an extension of matrices into three or more dimensions. The data that occur in a wide range of areas such as numerical analysis, computer vision, neuroscience and data mining are generally in the form of sparse tensors. Tensor factorization is the process of analysis that is performed in order to discover the latent features of these multi-dimensional data. In this work, we propose a new technique for distributed sparse tensor factorization by an intelligent partitioning of work load into processors. The experimental results on real-world sparse tensors verifies the effectiveness of the proposed model.


DATE: 10 October, 2016, Monday @ 17:00