delo diplomskega seminarja
Sara Kovačič (Author), Aljoša Peperko (Mentor)

Abstract

V delu je predstavljena regresija z Gaussovimi procesi iz vidika uteženega prostora in s pogledom iz prostora funkcij. Ponovljenih je nekaj osnov Bayesove statistike in lastnosti normalne porazdelitve. Za namene modeliranja in strojnega učenja je predstavljena tudi teorija učenja modela. Ker so z Gaussovimi procesi tesno povezane kovariančne funkcije, je predstavljenih nekaj najpogostejših kovariančnih funkcij. V empiričnem delu naloge sta opisana Pythonova knjižnica za strojno učenje Scikit-learn in primer regresije z Gaussovimi procesi na rezultatih nacionalnega preverjanja znanja za osnovnošolce iz leta 2019.

Keywords

matematika;Gaussovi procesi;kovariančne funkcije;regresija;strojno učenje;

Data

Language: Slovenian
Year of publishing:
Typology: 2.11 - Undergraduate Thesis
Organization: UL FS - Faculty of Mechanical Engineering
Publisher: [S. Kovačič]
UDC: 519.2
COBISS: 18737241 Link will open in a new window
Views: 1395
Downloads: 264
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: Gaussian process regression
Secondary abstract: The thesis presents the Gaussian process regression from the weight space view and the function space view. It examines some of Bayesian statistics and normal distribution properties. For modeling and machine learning purposes the model learning theory is also presented. Since covariance functions are tightly connected to the Gaussian process the thesis contains a presentation of the most frequent covariance functions. The empirical part of the thesis includes a description of Python’s Scikit-learn machine learning library as well as an example of the Gaussian process regression based on the results of the 2019 national assessment of elementary school students in Slovenia.
Secondary keywords: mathematics;Gaussian processes;covariance functions;regression;machine learning;
Type (COBISS): Final seminar paper
Study programme: 0
Embargo end date (OpenAIRE): 1970-01-01
Thesis comment: Univ. v Ljubljani, Fak. za matematiko in fiziko, Oddelek za matematiko, Finančna matematika - 1. stopnja
Pages: 28 str.
ID: 11229752