Neuronske mreže izgrađene maxout-neuronima

undergraduate thesis

• Title (english): Maxout Neural Networks
• Author: Luka Žmegač
• Mentor: Marko Čupić (mentor)
• Committee member: Marko Čupić (predsjednik povjerenstva)
• Committee member: Davor Kukolja (član povjerenstva)
• Committee member: Alan Jović (član povjerenstva)
• Granter: University of Zagreb; Faculty of Electrical Engineering and Computing; Zagreb
• Defense date and country: 2016-07-11, Croatia
• Scientific / art field, discipline and subdiscipline: TECHNICAL SCIENCES; Computing

Abstract

Umjetne neuronske mreže danas pospješuju napredak umjetne inteligencije. Njihova primjena je prisutna u velikoj većini industrija gdje rješavaju različite probleme. Neuronske mreže nisu primjenjive u svim situacijama. Maxout-neuroni koji su obrađeni u ovom radu, nova su vrsta neurona koja uz tehniku Dropout pokušavaju riješiti određene probleme mreža. Napravljena je implementacija neuronske mreže tipa Maxout i algoritma propagacije pogreške unatrag uz korištenje tehnike Dropout. Mreža je istestirana na raznim problemima regresije i klasifikacije gdje je i uz manju kompleksnost pokazala vrlo dobre rezultate.

Abstract (english)

Nowadays artificial neural networks foster great advancements in field of Artificial Inteligence. Its application can be seen in most industries where is employed to solve diverse problems, but still there are some for which neural networks are not working. Maxout neurons which are researched in this work are new type of neurons which in association with Dropout training try to solve some of the problems other types of neural networks encounter. In this work Maxout neural networks and Backpropagation algorithm enhanced with Dropout technique have been implemented. Neural networks have been tested on both regression and classification problems and have achieved great results even with little network complexity.

• Keywords: Neuronske mreže; Maxout; neuroni; Dropout; algoritam propagacije pogreške unatrag; univerzalni aproksimator
• Keywords (english): Neural networks; Maxout; neuron; Dropout; Backpropagation algorithm; universal aproximator
• Language: croatian
• URN:NBN: urn:nbn:hr:168:232398
• Study programme: Title: Computing; Study programme type: university; Study level: undergraduate; Academic / professional title: sveučilišni/a prvostupnik/ prvostupnica (baccalaureus/baccalaurea) inženjer/inženjerka računarstva (sveučilišni/a prvostupnik/ prvostupnica (baccalaureus/baccalaurea) inženjer/inženjerka računarstva)
• Type of resource: Text
• File origin: Born digital
• Access conditions: Closed access
• Created on: 2018-02-06 14:37:44