Softverska implementacija i optimizacija hardverske neuronske mreže za raspoznavanje slova

undergraduate thesis

• Title (english): Software implementation and optimization of a hardware neural network for letter recognition
• Author: Ivan Vlašić
• Mentor: Mirko Poljak (mentor)
• Committee member: Mirko Poljak (predsjednik povjerenstva)
• Committee member: Vladimir Čeperić (član povjerenstva)
• Committee member: Mislav Matić (član povjerenstva)
• Granter: University of Zagreb; Faculty of Electrical Engineering and Computing; Zagreb
• Defense date and country: 2023-06-15, Croatia
• Scientific / art field, discipline and subdiscipline: TECHNICAL SCIENCES; Computing

Abstract

Neuronske mreže danas se sve više koriste u različitim dijelovima znanosti, posebice u području raspoznavanja uzoraka. U ovom radu napravljena je softverska implementacija hardverske mreže realizirane na čipu iz literature koja raspoznaje crno bijele slike slova niske rezolucije (A, V, T i X). Neuronska mreža trenirana je i validirana na 40 uzoraka, te testirana na 640. Preciznost ovakve neuronske mreže dostiže 84%, što pokazuje da može poslužiti kao simulacija hardverskom rješenju. Predloženo poboljšanje u vidu povećanja broja neurona maksimalno 2 puta povećava brzinu učenja i diže preciznost na 91%.

Abstract (english)

Neural networks are increasingly used today in various parts of science, especially in the field of pattern recognition. In this thesis, a software implementation of a hardware network realized on a chip that recognizes black and white low-resolution images (letters A, V, T and X) is done. The neural network was trained and validated on 40 samples, then tested on 640 samples. Accuracy of this neural network reaches 84%, which makes it a useful tool for the simulation for the hardware implementation. Proposed improvement in the form of increasing the number of neurons by a maximum of 2 times increases the learning speed and accuracy (from 84% to 91%).

• Keywords: neuronska mreža; višeslojni perceptron; raspoznavanje uzoraka; memristori; neuronska mreža na čipu
• Keywords (english): neural network; multilayer perceptron; pattern recognition; memristors; on-chip neural network
• Language: croatian
• URN:NBN: urn:nbn:hr:168:278232
• 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: 2023-10-11 10:55:59