prikaz prve stranice dokumenta Algoritmi poduprtog učenja i dilema istraživanja i iskorištavanja
Access restricted to students and staff of home institution
master's thesis
Algoritmi poduprtog učenja i dilema istraživanja i iskorištavanja
2020. urn:nbn:hr:166:047345
Šušić, Igor
University of Split
University of Split, Faculty of science
Department of Informatics

Cite this document

Šušić, I. (2020). Algoritmi poduprtog učenja i dilema istraživanja i iskorištavanja (Master's thesis). Retrieved from https://urn.nsk.hr/urn:nbn:hr:166:047345

Šušić, Igor. "Algoritmi poduprtog učenja i dilema istraživanja i iskorištavanja." Master's thesis, University of Split, University of Split, Faculty of science, 2020. https://urn.nsk.hr/urn:nbn:hr:166:047345

Šušić, Igor. "Algoritmi poduprtog učenja i dilema istraživanja i iskorištavanja." Master's thesis, University of Split, University of Split, Faculty of science, 2020. https://urn.nsk.hr/urn:nbn:hr:166:047345

Šušić, I. (2020). 'Algoritmi poduprtog učenja i dilema istraživanja i iskorištavanja', Master's thesis, University of Split, University of Split, Faculty of science, accessed 21 May 2022, https://urn.nsk.hr/urn:nbn:hr:166:047345

Šušić I. Algoritmi poduprtog učenja i dilema istraživanja i iskorištavanja [Master's thesis]. Split: University of Split, University of Split, Faculty of science; 2020 [cited 2022 May 21] Available at: https://urn.nsk.hr/urn:nbn:hr:166:047345

I. Šušić, "Algoritmi poduprtog učenja i dilema istraživanja i iskorištavanja", Master's thesis, University of Split, University of Split, Faculty of science, Split, 2020. Available at: https://urn.nsk.hr/urn:nbn:hr:166:047345

Please login to the repository to save this object to your list.
Metadata
TitleAlgoritmi poduprtog učenja i dilema istraživanja i iskorištavanja
Title (english)Reinforcement learning algorithms and exploration exploitation dilemma
AuthorIgor Šušić
MentorSaša Mladenović (mentor)
Committee memberSaša Mladenović (predsjednik povjerenstva)
Committee memberGoran Zaharija (član povjerenstva)
Committee memberDivna Krpan (član povjerenstva)
GranterUniversity of Split
University of Split, Faculty of science
(Department of Informatics)
Split
Defense date and country2020-09-17, Croatia
Scientific / art field,
discipline and subdiscipline		TECHNICAL SCIENCES
Computing
Abstract
Ovaj rad nudi pregled poduprtog učenja, ideja i algoritama na kojima se područje temelji. Kroz primjere polazi se od formalizacije problema koristeći Markovljeve procese odluke sve do Q-učenja. Promatra dilemu istraživanja i iskorištavanja, daje poveznicu s ponašanjem živih bića. Analizira te uspoređuje uspjeh različitih politika ponašanja kao rješenja dileme, točnije uspoređuje pohlepni epsilon, softmax i gornju granicu intervala pouzdanosti nad jednostavnim labirintom. Softmax je jedina politika koja u ograničenom broju epizoda konvergira na optimalnu politiku ponašanja za primjer korišten u radu.
Abstract (english)
This work gives an overview of reinforcement learning, ideas, and algorithms that it is based upon. Through examples, it goesfrom formalization using MDP to Q-learning. Observesthe explorationexploitation dilemma and gives a link to the psychology of living creatures. Analyzes and compares multiple policies, such as epsilon greedy, softmax, and upper confidence bound in a use case of a simple maze. In such envirovment with some restraints like finite episode count only softmax converged to the optimal policy.
Keywords
Keywords (english)
Languagecroatian
URN:NBNurn:nbn:hr:166:047345
Study programmeTitle: Computer Science; specializations in: major in teaching
Course: major in teaching
Study programme type: university
Study level: graduate
Academic / professional title: magistar/magistra edukacije informatike (magistar/magistra edukacije informatike)
Type of resourceText
File originBorn digital
Access conditionsAccess restricted to students and staff of home institution
Terms of useLicence In copyright icon
Created on2021-01-13 16:54:19