diplomsko delo
Bian Klančnik (Author), Matjaž Kukar (Mentor)

Abstract

V zadnjem desetletju se je strojno učenje precej razvilo in prodira na vsa področja informacijskih tehnologij. Dandanes večina računalniških sistemov uporablja strojno učenje na tak ali drugačen način. Poleg tega pa je zelo napredovalo tudi strojno učenje na manj zmogljivih napravah. Cilj diplomske naloge je preizkusiti učinkovitost obstoječih orodij za strojno učenje na manj zmogljivih napravah. Osredotočili smo se na naprave ARM. Na osebnem računalniku smo zgradili več različnih modelov v različnih ogrodjih za grajenje modelov strojnega učenja. Modele smo serializirali s pomočjo orodij za serializacijo in jih na koncu pognali na Raspberry Pi. Zgradili smo več klasifikacijskih in en regresijski model. Merili smo uspešnost modelov in čas, ki ga model na določeni napravi porabi za napovedovanje. Rezultati so pokazali, da se uspešnost modelov na različnih napravah ne razlikuje. Razlika v izmerjenem času pa se je med napravami precej razlikovala.

Keywords

vgrajene naprave;serializacija modelov;visokošolski strokovni študij;diplomske naloge;

Data

Language: Slovenian
Year of publishing:
Typology: 2.11 - Undergraduate Thesis
Organization: UL FRI - Faculty of Computer and Information Science
Publisher: [B. Klančnik]
UDC: 004.85(043.2)
COBISS: 77621507 Link will open in a new window
Views: 280
Downloads: 25
Average score: 0 (0 votes)
Metadata: JSON JSON-RDF JSON-LD TURTLE N-TRIPLES XML RDFA MICRODATA DC-XML DC-RDF RDF

Other data

Secondary language: English
Secondary title: Machine learning on embedded platforms
Secondary abstract: Machine learning has developed considerably in the last decade and is penetrating all areas of information technology. Today, most computer systems use machine learning in one way or another. In addition, machine learning on less powerful devices has advanced greatly. The aim of the diploma thesis is to test the effectiveness of existing machine learning tools on less powerful devices. We focused on ARM devices. We built several different models on a personal computer in different frameworks to build machine learning models. We serialized the models using serialization tools and eventually ran them on a Raspberry Pi. We built several classification and one regression model. We measured the performance of the models and the time that the model spends on a particular device to predict. The results showed that the performance of the models on different devices did not differ. The difference in measured time, however, varied considerably between devices.
Secondary keywords: machine learning;embedded devices;model serialization;computer science;computer and information science;diploma;Strojno učenje;Umetna inteligenca;Računalništvo;Univerzitetna in visokošolska dela;
Type (COBISS): Bachelor thesis/paper
Study programme: 1000470
Thesis comment: Univ. v Ljubljani, Fak. za računalništvo in informatiko
Pages: 52 str.
ID: 13394711