COMPUTER SIMULATION OF NEURAL NETWORK OPERATION FOR OBJECT RECOGNITION

Abstract

Aim. We construct a computer model of a neural network for object recognition. Methodology. Based on the ideas underlying the theory of recognition and the theory of neural networks, we have constructed a model of neural network operation designed to recognize the studied mappings with a given accuracy. For the successful operation of a neural network, we have used open access databases from remote servers, which makes it possible to use neural networks in cramped conditions (in the absence of powerful computers). A program in Python has been developed to organize and manage the neural network. Results. A neural network has been constructed that recognizes the studied mappings with a given accuracy. To manage the built neural network and attract database arrays from remote servers, a program in Python has been developed. The principle of neural network operation is demonstrated in practice, using the example of recognition of the studied mappings. Research implications. The model provides a real recipe for constructing a neural network and using it in practice in the absence of a powerful computer.