Implementing dynamical pattern recognition algorithm on computer cluster

Abstract

Dynamical pattern recognition based on deterministic learning is the latest pattern recognition algorithm. To accomplish large-scale calculation in a short time under this algorithm, high-efficient computing platform is needed. Parallel computing and computer cluster provide an efficient method of solving this problem. In this paper, the parallel implementation of dynamical pattern recognition algorithm is accomplished firstly, and then a computer cluster based on MATLAB for this recognition algorithm is built. In the end, an experiment about the detection of rotating stall in an axial compressor through dynamic recognition on the cluster is included to demonstrate the effectiveness of the computer cluster. It shows that based on the computer cluster, the parallelization technique can dramatically improve the efficient of the recognition algorithm.