Application of genetic algorithm for selection of neyral network hyperparameters

Machine learning and knowledge control systems
Authors:
Platonov V. V. Yanbarisova Ya. E.
Abstract:

The widespread use of various neural networks for detecting cyberattacks is hindered by the difficulty of determining their hyperparameters. Typically, hyperparameter values are established experimentally. This paper presents an approach to selecting perceptron hyperparameters for network attack detection using a genetic algorithm. Experimental results confirm the validity of this approach.