Unit:
Κατεύθυνση Στατιστική και Επιχειρησιακή ΈρευναLibrary of the School of Science
Author:
Λογοθέτης Δημήτριος
Supervisors info:
Σάμης Τρέβεζας Λέκτορας
Original Title:
Αλγοριθμικές τεχνικές μπεϋζιανής και κλασική προσέγγισης σε μοντέλα ανάπτυξης φυτών και ζητήματα σύγκλισης στο σύνορο του παραμετρικού χώρου
Translated title:
Algorithmic techniques of bayesian and classical approach in plant growth models and convergence issues in the boundary of the parameter space
Summary:
In this thesis, two problems were studied which arize in the framework of
plant growth models. The first one is a theoretical issue concerning the
statistical method Gaussian Randomization and more specifically the
convergence of an EM algorithm in order to approximate the solution. The
desirable convergence point falls into the boundary of parameter space creating
a non typical situation. The second one is a parameter estimation problem in
the Greenlab model and an algorithm of bayesian approach was designed. The
results of which, were compared, with the results of previous algorithmics
techniques of classical approach.
Keywords:
Hidden Markov Models, EM and GEM algorithm, Plant growth models, Maximum likelihood estimator, Bayesian approach
