Εντοπίστηκε ένα σφάλμα στη λειτουργία της ΠΥΞΙΔΑΣ όταν χρησιμοποιείται μέσω του προγράμματος περιήγησης Safari. Μέχρι να αποκατασταθεί το πρόβλημα, προτείνουμε τη χρήση εναλλακτικού browser όπως ο Chrome ή ο Firefox. A bug has been identified in the operation of the PYXIDA platform when accessed via the Safari browser. Until the problem is resolved, we recommend using an alternative browser such as Chrome or Firefox.
Λογότυπο αποθετηρίου

Μεταπτυχιακές Εργασίες

Μόνιμο URI για αυτήν τη συλλογήhttps://pyxida.aueb.gr/handle/123456789/7

Περιήγηση

Πλοήγηση Μεταπτυχιακές Εργασίες ανά Επιβλέπων "Demiris, Nikolaos"

A B C D E F G H I J K L M N O P Q R S T U V W X Y Z
Α Β Γ Δ Ε Ζ Η Θ Ι Κ Λ Μ Ν Ξ Ο Π Ρ Σ Τ Υ Φ Χ Ψ Ω
Φιλτράρετε τα αποτελέσματα πληκτρολογώντας τα πρώτα γράμματα
Τώρα δείχνει 1 - 1 από 1
  • Μικρογραφία εικόνας
    Τεκμήριο
    Statistical and computational efficiency comparison between MCMC and Hamiltonian Monte Carlo with application in epidemiological models
    (2018) Barmpounakis, Petros; Μπαρμπουνάκης, Πέτρος; Demiris, Nikolaos; Athens University of Economics and Business, Department of Informatics
    In the recent years several methods have been proposed and applied to different data problems inorder to perform Bayesian inference for the posterior density of the model’s parameters. Fornearly three decades the most widely used are Markov Chain Monte Carlo methods andspecifically Metropolis-Hastings and Gibbs Sampling, which will be examined for the purposesof this thesis. The new hot topic in Bayesian Inference is Hamiltonian Monte Carlo, whichpromises respectable results as regards accuracy and speed. In epidemiology it is even morecrucial to find an algorithm that returns the most accurate results in the most efficient way. Inthis project, the two aforementioned algorithms will be implemented in a temporal-stochasticepidemiology model regarding the spreading of the Sheeppox virus in the region of Evros,Greece. The two algorithms will be compared by their computational efficiency based on theCPU time and their statistical efficiency after comparing the results returned from the prequentialanalysis, using appropriate scoring rules. At this point, it is of crucial importance to highlight thedifferent softwares used to implement the algorithms; Metropolis-Hastings and Gibbs Samplingrun through WinBUGS and Hamiltonian Monte Carlo through the new probabilistic languageStan. Consequently, any differences occurring in the results may also be derivatives of thedifferent softwares usage. At this stage, it is important to mention that Stan is on experimentallevel; hence some inaccuracies in the results may occur. The interface for both softwares ischosen to be R.