Skladanje glazbe pomoću rekurzivnih neuronskih mreža

undergraduate thesis

• Title (english): Composing music with recurrent neural networks
• Author: Jure Kostelac
• Mentor: Siniša Srbljić (mentor)
• Committee member: Siniša Srbljić (predsjednik povjerenstva)
• Committee member: Adrian Satja Kurdija (član povjerenstva)
• Committee member: Miroslav Popović (član povjerenstva)
• Granter: University of Zagreb; Faculty of Electrical Engineering and Computing; Zagreb
• Defense date and country: 2018-09-05, Croatia
• Scientific / art field, discipline and subdiscipline: TECHNICAL SCIENCES; Computing

Abstract

U ovom radu obrađuje se metoda skladanja glazbe pomoću rekurzivnih neuronskih mreža. Na početku prolazimo tijek razvoja neuronskih mreža. Uspoređujemo biološki neuron s umjetnim, navodimo vrste prijenosnih funkcija i govorimo o usponima i padovima neuronskih mreža. Nadalje objašnjavamo razliku između unaprijednih i ponavljajućih mreža. Spominjemo algoritam propagacije pogreške unatrag koji je zaslužan za rješavanje problema dodjele zasluga. Skladanje glazbe radimo na primjeru dvoosne ponavljajuće neuronske mreže. Analiziramo MIDI datoteke iz kojih izvlačimo elemente koje vrednujemo i objašnjavamo kako se ostvaruje vremenska i notna invarijantnost neuronske mreže. Rad zaključujemo uspoređujući prednosti i mane ove metode.

Abstract (english)

In this bachelor thesis we process a method of composing musical works using recursive neural networks. First we go through the history of neural networks. We compare biological neurons with artificial, we refer the types of transmission functions and talk about the rise and fall of neural networks. We further explain the difference between feed-forward and recurring networks. Backpropagation algorithm is mentioned as the algorithm responsible for solving the Credit Assignment Problem. We compose music using a biaxial recurrent neural network. We analyze MIDI files from which we draw elements that we value and explain how to achieve time and note invariance. We conclude by comparing the advantages and disadvantages of this method.

• Keywords: umjetna inteligencija; skladanje glazbe; umjetna neuronska mreža
• Keywords (english): artificial intelligence; composing music; artificial neural network
• Language: croatian
• URN:NBN: urn:nbn:hr:168:036777
• 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: 2019-03-13 21:44:56