The influence of hyperparameters on the performance of deep learning models

Tema ovog rada je utjecaj hiperparametara na rad modela dubokog učenja. Hiperparametri predstavljaju parametre koji kontroliraju proces učenja. Kao model dubokog učenja odabrana je duboka konvolucijska neuronska mreža koja predstavlja temelj algoritama računalnog vida. Rad se sastoji od teorijskog i praktičnog dijela. U sklopu teorijskog dijela iznesena je sva teorija potrebna za razumijevanje utjecaja pojedinih hiperparametara na rad modela dubokog učenja. Praktični dio rada sastoji se od izrade modela duboke konvolucijske neuronske mreže i podešavanja njegovih hiperparametara u programskom jeziku Python.The theme of this work is the influence of hyperparameters on the performance of the deep learning models. Hyperparameters represent parameters that control the learning process. A deep convolution neural network is selected as a deep learning model that forms the basis of computer vision algorithms. The work consists of a theoretical and practical part. The theoretical part sets out all the theories necessary to understand the influence of individual hyperparameters on the performance of the deep learning models. The practical part of the work consists of developing a model of a deep convolution neural network and setting up its hyperparameters in the Python programming language