Recognition of multiple sequences by subgroups of autonomous probabilistic automata

Abstract

Recognition of a group of sequences is carried out using a set of k subgroups, each of which includes b autonomous probabilistic automata (APA) set on the basis of ergodic stochastic matrices (ESM) of a certain subclass. Each of the b APA within k subgroups has a common set of states of a given power, and the vector of the initial distribution of each of these APA is determined by the distribution law of the first element of the corresponding sequence. The subgroup input from b APA receives element-wise members of b sequences.. The obtained probability vectors are subjected to clustering by the k-means method, which is based on the previously proposed methodology for a sample of reference object introduction. The clustering technique makes it possible to identify the groups of sequences by determining the probability of identification at each step of the algorithm, which can be implemented in parallel when each sequence is identified from a given set by each APA, parallelizing this process.