Stan v oblaku

magistrsko delo

Uroš Hercog (Author), Erik Štrumbelj (Mentor), Matjaž Pančur (Co-mentor)

Abstract

Zaradi vse hitrejšega razvoja in dostopnosti zmogljive strojne opreme je računalništvo v oblaku postalo zanimivo na področju dela s podatki, strojnega učenja in statistike. Izvajanje učenja modelov in obdelave veliko podatkov želijo uporabniki premakniti v visoko-stopnjevalne oblake. Tam te procese izvedejo z višjo stopnjo vzporednega izvajanja, kot jo lahko dosežejo na osebnih računalnikih. Vendar pa so uporabniki omejeni pri uporabi orodij, saj vsa oddaljenega izvajanja ne podpirajo. Takšno orodje je tudi Stan, za katerega smo v nalogi razvili rešitev v oblaku. Pregledali smo področje orodij statističnega modeliranja, poiskali sorodne rešitve in obstoječe rešitve, ki omogočajo uporabo orodja Stan v oblaku. Zbrali smo funkcionalne zahteve in razdelali arhitekturo platforme za Stan v oblaku. Na podlagi predlagane arhitekture smo razvili rešitev, imenovano Cloudstan. Platforma je sestavljena iz zalednega dela, ki skrbi za orkestracijo izvajanja prevajanja in vzorčenja modelov Stan, orodje ukazne vrstice in knjižnico v programskem jeziku R, ki omogočata komunikacijo z zalednim sistemom.

Keywords

Saas;Stan;Bayesova statistika;računalništvo in informatika;magisteriji;

Data

Language:	Slovenian
Year of publishing:	2021
Typology:	2.09 - Master's Thesis
Organization:	UL FRI - Faculty of Computer and Information Science
Publisher:	[U. Hercog]
UDC:	004.76(043.2)
COBISS:	84027139
Views:	264
Downloads:	31
Average score:	0 (0 votes)
Metadata:

Other data

Secondary language:	English
Secondary title:	Bayesian statistics in the cloud
Secondary abstract:	The proliferation and accessibility of high-performance hardware made cloud computing interesting in the areas of data processing, machine learning and statistics. Users are moving the model training and processing of data to scalable cloud solutions which allow them to execute these processes in a highly parallel manner. This allows them to complete their tasks in less time than on personal computers. But not all tools used by experts have native support for remote execution. In this master’s thesis, we developed a cloud solution for a tool for statistical modeling called Stan. We analyzed and compared cloud solutions for tools similar to Stan. We collected functional requirements and presented the system architecture. Based on the architecture, we developed the platform called Cloudstan, a command-line interface and a library for communicating with the platform written in R.
Secondary keywords:	SaaS;Stan;Bayessian statistics;cloud computing;computer science;computer and information science;master's degree;Računalništvo v oblaku;Računalništvo;Univerzitetna in visokošolska dela;
Type (COBISS):	Master's thesis/paper
Study programme:	1000471
Thesis comment:	Univ. v Ljubljani, Fak. za računalništvo in informatiko
Pages:	51 str.
ID:	13805882