Primjena evolucjiskog algoritma za učenje neuronske mreže

master's thesis

• Title (english): Application of an evolutionary algorithm for neural network learning
• Author: Krešimir Medovka
• Mentor: Petar Ćurković (mentor)
• Committee member: Damir Markučič (predsjednik povjerenstva)
• Committee member: Darko Landek (član povjerenstva)
• Committee member: Petar Ćurković (član povjerenstva)
• Granter: University of Zagreb; Faculty of Mechanical Engineering and Naval Architecture; Zagreb
• Defense date and country: 2022-07-18, Croatia
• Scientific / art field, discipline and subdiscipline: TECHNICAL SCIENCES; Computing; Artificial Intelligence

Abstract

Tema ovog diplomskog rada bila je Primjena evolucijskog algoritma za učenje neuronske mreže. Opisane su osnove neuronskih mreža i njihove karakteristike kao što su čvorovi i slojevi, te vrste neuronskih mreža. Također opisani su evolucijski algoritmi, inspiracija, struktura algoritma i potrebni koraci. Dan je opis i pregled područja NeuroEvolucije s naglaskom na Neuronske mreže evoluirajućih topologija i težina (eng. Topology and Weight Evolving Artificial Neural Networks - TWEANN) te su postavljena neka vrlo bitna pitanja koje će NEAT algoritam pokušati riješiti. Proučen je teorijski koncept NeuroEvolucije promjenjivih topologija (eng. NeuroEvolution of Augmenting Topologies - NEAT). NEAT algoritam je implementiran u programskom okruženju MATLAB te je implementacija ispitana na dva problema: XOR problem i problem stabilizacije inverznog njihala.

Abstract (english)

The topic of this thesis was Application of an evolutionary algorithm for neural network learning. The basics of neural networks and their characteristics such as nodes and layers, and types of neural networks are described. Evolutionary algorithms, inspiration, algorithm structure and necessary steps are also described. A description and overview of the field of NeuroEvolution with an emphasis on Topology and Weight Evolving Artificial Neural Networks (TWEANN) was given, and some very important questions were asked that the NEAT algorithm will try to solve. The theoretical concept of NeuroEvolution of Augmenting Topologies (NEAT) was studied. The NEAT algorithm was implemented in the MATLAB programming environment, and the implementation was tested on two problems: the XOR problem and the inverse pendulum stabilization problem.

• Keywords: NEAT; neuronske mreže; evolucijski algoritmi
• Keywords (english): NEAT; neural networks; evolutionary algorithms
• Language: croatian
• URN:NBN: urn:nbn:hr:235:466283
• Study programme: Title: Mechanical Engineering; Study programme type: university; Study level: graduate; Academic / professional title: magistar/magistra inženjer/inženjerka strojarstva (magistar/magistra inženjer/inženjerka strojarstva)
• Type of resource: Text
• File origin: Born digital
• Access conditions: Open access
• Created on: 2022-07-17 21:15:49