Neural network structure optimization tool

This thesis deals with optimizing the structures of artificial and convolutional neural networks. The hyperparameters, from which these structures are comprised of, are described in the theoretical part of this thesis. In addition, it explains the metrics used for evaluation of these structures. The practical outcome of this thesis is a tool capable of automatically generating neural network structures for a given dataset based on userdefined configuration. The tool also automatically tests the generated structures and creates reports which summarize the performace of the best generated structures. The tool is implemented using Python language, with utilization of TensorFlow and Keras libraries. In addition to providing a detailed source code description, the practical part of the thesis includes testing the tool on well-known datasets, as well as a dataset simulating traffic of an industrial network under ongoing cyber attack