Uporaba evolucijskih algoritama u podrezivanju konvolucijskih neuronskih mreža

master's thesis

• Title (english): Evolutionary algorithms for pruning of convolutional neural networks
• Author: Filip Škrlec
• Mentor: Domagoj Jakobović (mentor)
• Committee member: Domagoj Jakobović (predsjednik povjerenstva)
• Committee member: Alan Jović (član povjerenstva)
• Committee member: Nikolina Frid (član povjerenstva)
• Granter: University of Zagreb; Faculty of Electrical Engineering and Computing; Zagreb
• Defense date and country: 2024-07-09, Croatia

Abstract

U ovom radu implementiran je algoritam temeljen na evolucijskim algoritmima koji je korišten za proces optimizacije obrezivanja konvolucijskih neuronskih mreža. Uz opis razvijanja i implementacije algoritma, dan je uvid u teorijsku pozadinu rada evolucijskih algoritama i konvolucijskih neuronskih mreža. Opisane su trenutne tehnike obrezivanja konvolucijskih neuronskih mreža te je na implementiranom algoritmu provedeno ispitivanje. Početna ispitivanja korištena su za optimizaciju hiperparametara algoritma dok su na kraju rada rezultati uspoređeni s relevantnim algoritmima i njihovim rezultatima.

Abstract (english)

In this paper, an algorithm based on evolutionary algorithms was implemented, which was used for the process of optimizing the pruning of convolutional neural networks. Along with the description of the development and implementation of the algorithm, an insight into the theoretical background of the work of evolutionary algorithms and convolutional neural networks is given. The current techniques of pruning convolutional neural networks are described, and the implemented algorithm is tested. Initial tests were used to optimize the hyperparameters of the algorithm, while at the end of the work the results were compared with relevant algorithms and their results.

• Keywords: evolucijski algoritmi; optimizacija; konvolucijske neuronske mreže; obrezivanje
• Keywords (english): evolutionary algorithms; optimization; convolutional neural networks; pruning
• Language: croatian
• URN:NBN: urn:nbn:hr:168:170086
• Study programme: Title: Computing; Study programme type: university; Study level: graduate; Academic / professional title: magistar/magistra inženjer/inženjerka računarstva (magistar/magistra inženjer/inženjerka računarstva)
• Type of resource: Text
• File origin: Born digital
• Access conditions: Closed access
• Created on: 2024-10-12 22:16:49