Klasifikacija rukom pisanih znamenki upotrebom spektralne analize na grafu

undergraduate thesis

• Title (english): Classification of handwritten characters using spectral graph theory
• Author: Suzana Pratljačić
• Mentor: Marko Bosiljevac (mentor)
• Committee member: Marko Bosiljevac (predsjednik povjerenstva)
• Committee member: Sonja Grgić (član povjerenstva)
• Committee member: Dario Bojanjac (član povjerenstva)
• Granter: University of Zagreb; Faculty of Electrical Engineering and Computing; Zagreb
• Defense date and country: 2019-07-12, Croatia
• Scientific / art field, discipline and subdiscipline: TECHNICAL SCIENCES; Computing

Abstract

Spektralno klasteriranje metoda je analize podataka čiji je cilj grupirati podatke prema zadanom kriteriju sličnosti. U radu je objašnjena teorijska pozadina spektralne analize grafa, svojstva Laplaceove matrice grafa te način prikazivanja slika u formi grafa sličnosti. Prepoznavanje znamenke na slici ostavreno je kao dodavanje nove točke u graf. Implementirani su algoritam grupiranja zasnovan na spektralnim metodama i algoritam dodavanja nove točke u graf zasnovan na Nystromovoj metodi. Analizirana je kvaliteta prepoznavanja znamenke u ovisnosti o različitim parametrima. Razvijeni sustav ispitan je na bazi rukom pisanih znamenki (MNIST baza).

Abstract (english)

Spectral clustering is a method for analysis of data whose goal is to group data according to a certain similarity criterion. The paper explains the theoretical background of spectral analysis of graphs, properties of Laplacian matrix of a graph and different ways of representing pictures in a form of similarity graph. Recognising a digit on a picture is realized as an addition of a new point in the graph. We implemented a grouping algorithm based on spectral methods, along with point addition algorithm, based on Nystrom method. We analysed the quality of digit recognition and its dependency on a number of different parameters. The system was tested on the database of handwritten digits (MNIST database).

• Keywords: Spektralno klasteriranje; Laplaceova matrica; Nystromova metoda
• Keywords (english): Spectral clustering; Laplacian matrix; Nystrom method
• Language: croatian
• URN:NBN: urn:nbn:hr:168:677818
• 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: 2020-01-07 21:33:26