Parallel Algorithms for Multidimensional Data Streams Analysis with Tensor Subspace Models

Abstract

In this paper the parallel models for processing of the multi-dimensional data streams are discussed. Stream analysis is performed by tensor models and a fitness measure. The method was tested on the problem of video shot detection showing good accuracy. In this paper efficient algorithms for tensor model construction and model update in the parallel processing framework are presented. Also the parallel version for the off-line stream processing is proposed.