The present invention relates to a method for searching for an optimal structure of a convolution neural network. The method for searching for an optimal structure of a convolution neural network using a genetic algorithm according to the present invention, comprises the steps of: generating an initial generation by initializing a convolution neural network structure, and generating chromosomes of the initial generation; generating and learning a convolution neural network for the chromosomes, and calculating and evaluating suitability in accordance with a predetermined suitability function; selecting a superior chromosome through the evaluated suitability; generating a next generation candidate group from the superior chromosome through chromosome hybridization; applying a variation to the next generation candidate group through a chromosomal variation; and repeating the evaluation step, the selection step, the step of generating the next generation candidate group and the application step a predetermined number of times.
展开▼