Abstracts

Diversity of the Fauna

The cottoid fish are the one of the most studied groups now. According to last data (Sideleva, 2001), the fauna of cottoid fish of Baikal includes 3 families, 12 genera and 34 species. Determination of patterns of fish up to a species in some cases is rather compound problem and requires of high qualification of a specialist. Therefore the creation of electronic determinants is an actual problem. The basis of a creating determinant will be morphological signs, constitution of seismic sensory system, and in the future there will be data of molecular-biological researches. At an initial stage the most essential tags used at determination of species are selected: 28 key signs and more than 120 their statuses.

The determinant is realized with usage of neuronic network technology [1]. The intellectual systems on the basis of synthetic neuronic networks decide problems of classification successfully. Diverse, more traditional approaches to the solution to this problem are known also, but they haven’t necessary flexibility outside of limited conditions. For this purpose the multi-ply network of direct spreading of a type multi-ply perseptron is selected. Now programme implementation of algorithm of training of this network is worked up. The training selection is shaped of, on the basis of which one the network will be trained. The trained network is supposed to be integrated in multimedia system of the atlas - determinant, where she will decide a problem of recognition of species of cottoid fish.

The literature
1. Neuronic computer engineering: the theory and practice, F. Wossermann, translation on Russian, J. À. Zuev, V. À. Tochenov, 1992.

Note. Abstracts are published in author's edition

|Home Page| |English Part|

Go to Home