V. I. Klyachkin, А. V. Rudinsky

Evolutionary problems are analyzed on the base of incorporation of dynamic causal description and probabilistic algorithms of reception and transformation of useful signals against the background of noise in hydroacoustic and hydrophysical channels of informational distribution. The problem of signal processing in the accidental sequence of observed count was considered, which is performed at the discreet moments of time over the observed vector in the form of hypothesis check. Among the multiple of approaches to the decision of the present problem two outermost ones are distinguished: one corresponds with consequent decision-making of problem of detection, classification and determination of current coordinates of object; and the parallel one, requires decision of the problem for all superthreshold marks simultaneously with subsequent selection of objects in accordance with the rule of logical alliance and organization of algorithms of filtering of feedback for the remaining subthreshold marks (i.e. performing of en-route analysis of all indicating array, for example, according to the scheme of Calman-Biucy). All other variants of approaches will be intermediate, without contributing of principal changes in the proposed scheme. Parallels were developed between neural nets and multichannel adaptive systems. These parallels allow to declare that modeling of artificial neural nets may be carried out also on the base of apparatus of the theory of complex technical systems, including the problems of direct observations of evolutional dynamics of systems by means of transformation of probabilistic structures of the evaluation of accidental fields of the source.