Post-graduate theses
Current Record: 15 of 127
|
Title |
Parameterizing molecular models via gaussian approximation methods |
Alternative Title |
Ανάπτυξη αδροποιημένων μοριακών μοντέλων με μεθόδους Gaussian παλινδρόμησης |
Author
|
Κυριέρη, Βασιλική
|
Thesis advisor
|
Χαρμανδάρης , Ευάγγελος
Καλλιγιαννάκη, Ευαγγελία
|
Reviewer
|
Μακράκης, Γεώργιος
Πανταζής, Γιάννης
|
Abstract |
In this thesis, we study Molecular Dynamics simulations, a scientific field that focuses
on the physical motion of atoms on a very small order of magnitude. Today,
many chemical and biological processes, such as protein interactions, can be modeled
through MD simulations with aim of analyzing their properties. The challenging part
of studying MD simulations of complex systems refers to an as-accurate-as-possible
prediction of the structure-property relationship at the microscopic level and the expensive
calculations of the dynamic quantities due to the wide range of length and time
scales. By decreasing the number of degrees of freedom, the new system can be used
with fewer variables. This method, known as Coarse-Graining, maps the atomistic
particles into mesoscopic particles such as ”superatoms”.
There exists a variety of methods to obtain the total force of the mesoscopic system,
either parametric or non-parametric models. In this thesis, the Gaussian process
regression model, a flexible non-parametric family of models capable of approximating
functions using relatively small data sets, is applied to a simple system, a methane
system, and its results are compared with two more straightforward approximations.
One with a parametric pair potential, the Lennard-Jones potential, and another with
the Linear B-splines representation. We learn the approximate force fields, with the
Force Matching criterion of loss, which minimizes the average distance between the
atomistic forces and the approximate CG forces.
|
Language |
English |
Subject |
Machine learning |
|
Μηχανική μάθηση |
|
Μοντελοποίηση μοριακών συστημάτων |
Issue date |
2023-03-17 |
Collection
|
School/Department--School of Sciences and Engineering--Department of Mathematics and Applied Mathematics--Post-graduate theses
|
|
Type of Work--Post-graduate theses
|
Permanent Link |
https://elocus.lib.uoc.gr//dlib/a/a/4/metadata-dlib-1679571326-442379-32146.tkl
|
Views |
449 |