Towards MapReduce Algorithms for the Higher Order-Singular Value Decomposition

The Higher-Order Singular Value Decomposition (HO-SVD) is the generalisation of the Singular Value Decomposition (SVD) from matrices to tensors. These decompositions have similar mathematical properties and many useful applications in science. Unfortunately, the computation of the SVD, and especially of the HO-SVD, represent computationally very expensive tasks for modern computers. The main purpose of this thesis is to demonstrate how the HO-SVD can be computed in a partially distributed manner using a MapReduce algorithm. In addition, a possible application of this technique in the field of Latent Semantic Analysis is demonstrated, where semantic structures in multidimensional data can be detected using the mentioned decompositions.