Kompleksnost skrivenih Markovljevih modela

Tepić, Martina (2015) Kompleksnost skrivenih Markovljevih modela. Diploma thesis, Faculty of Science > Department of Mathematics.

[img]
Preview
PDF
Language: Croatian

Download (573kB) | Preview
[img] Archive (dodatni materijali)
Language: Croatian

Download (212kB)

Abstract

This thesis is concerned with a statistical model called hidden Markov model (HMM), powerful statistical tool designed for modelling sequences generated by hidden processes. We gave a formal definition of the hidden Markov model, described various algorithms and implemented them in Python. Using the example of two asymmetric dies, we designed an HMM and used Viterbi training with and without deterministic annealing to optimize the parameters. We have noticed that Viterbi training with deterministic annealing works better than the one without annealing.

Item Type: Thesis (Diploma thesis)
Supervisor: Goldstein, Pavle
Date: 2015
Number of Pages: 26
Subjects: NATURAL SCIENCES > Mathematics
Divisions: Faculty of Science > Department of Mathematics
Depositing User: Iva Prah
Date Deposited: 14 Jul 2015 11:45
Last Modified: 14 Jul 2015 11:45
URI: http://digre.pmf.unizg.hr/id/eprint/4124

Actions (login required)

View Item View Item