Raspoznavanje uzoraka dubokim učenjem

master's thesis

• Author: Atina Koljevina
• Mentor: Luka Grubišić (mentor)
• Committee member: Luka Grubišić (predsjednik povjerenstva)
• Committee member: Robert Manger (član povjerenstva)
• Committee member: Igor Pažanin (član povjerenstva)
• Committee member: Nenad Antonić (član povjerenstva)
• Granter: University of Zagreb; Faculty of Science; (Department of Mathematics); Zagreb
• Defense date and country: 2017-02-22, Croatia
• Scientific / art field, discipline and subdiscipline: NATURAL SCIENCES; Mathematics

Abstract

U ovom radu detaljno su opisane duboke neuronske mreže, s posebnim naglaskom na povratnim neuronskim mrežama. Obrađene su metode i postupci koji se koriste kod treniranja dubokih neuronskih mreža. Kao primjer primjene opisana je izgradnja konverzacijskog agenta, zvanog chatbot. Implementiran je model chatbota u alatu TensorFlow temeljen na slijed-u-slijed arhitekturi i isproban je na dva skupa podataka: Cornell Movie-Dialogs Corpus skupu podataka i Twitter skupu podataka. Na kraju su prikazani rezultati evaluacije zajedno s komentarima.

Abstract (english)

This work describes deep neural networks, with emphasis on recurrent neural networks. Gradient methods and learning algorithms used for neural networks training are shown. As an application we show an implementation of a particular type of conversational agent, called the chatbot. Sequence-to-sequence model for conversational modeling problem is built using TensorFlow, and its performance is evaluated on two datasets: Cornell Movie-Dialogs Corpus dataset and Twitter dataset. Finally, first results of an evaluation are presented together with a discussion of the results.

• Keywords: duboke neuronske mreže; povratne neuronske mreže; treniranje dubokih neuronskih mreža; konverzacijski agent; chatbot; TensorFlow
• Keywords (english): deep neural networks; recurrent neural networks; neural networks training; conversational agent; chatbot; TensorFlow
• Language: croatian
• URN:NBN: urn:nbn:hr:217:879585
• Study programme: Title: Computer Science and Mathematics; Study programme type: university; Study level: graduate; Academic / professional title: magistar/magistra računarstva i matematike (magistar/magistra računarstva i matematike)
• Type of resource: Text
• File origin: Born digital
• Access conditions: Access restricted to students and staff of home institution
• Created on: 2017-04-26 11:32:40