FIELD: computer technology.
SUBSTANCE: invention relates to smart controllers using a genetic algorithm for search for a neural network structure and weight coefficients of its synoptical connections. A device contains an object of control, a block for training a neural network, a block for assessment of neural network performance, a data block, a neural network block, a block of a random number generator, a control block, a hybridization block, a mutation block, a chromosome history block, a chromosome selection block.
EFFECT: increase in the speed of finding a suboptimal neural network satisfying set indicators of the quality of operation.
1 cl, 1 dwg
Authors
Dates
2022-11-29—Published
2021-04-21—Filed