Your browser does not support JavaScript!

Home    Search  

Results - Details

Search command : Author="Ζέζας"  And Author="Ανδρέας"

Current Record: 1 of 19

Back to Results Previous page
Next page
Add to Basket
[Add to Basket]
Identifier 000425704
Title Classification of X-Ray binary systems using machine learning methods
Alternative Title Ταξινόμηση των δυαδικών συστημάτων ακτίνων-Χ με τη χρήση μεθόδων μηχανικής μάθησης
Author Μπέρτσιας, Αντώνιος Κ.
Thesis advisor Ζέζας, Ανδρέας
Reviewer Κυλάφης, Νικόλαος
Reig, Pablo
Abstract X-ray binaries (XRBs) are stellar systems which consist of a stellar remnant and a ‘normal’ (or donor) star. The two objects, orbiting around their common center of mass, are close enough to interact and exchange mass with the compact object accreting mass from the donor star. The energy released in XRBs is produced through the accretion of this material onto the compact object. Nearly all BHBs are transient X-ray sources that are discovered during an outburst. These variations in X-ray intensity led to the concept of X-ray states. The aim of this work is to develop two-dimensional and three-dimensional diagnostic diagrams for the characterization of black hole X-Ray binaries accretion states and classify the type of compact objects (black hole or neutron star) for observations with the X-ray observatory NuSTAR. Our sample consists of characterized RXTE spectra (2685 from 34 neutron-star Low Mass X-Ray Binary (LMXRB) sources, and 8976 from 14 BHXBs), which have been processed so as to simulate the spectra that would have been obtained using the NuSTAR X-Ray observatory. The two hardness ratios (namely HR1 and HR2) and the count rate were used in order to develop diagnostic maps for the classification of unlabeled X-ray binaries. The optimum separating hyperplanes were generated with the use of Support Vector Machine algorithm (SVM). The Kernel functions that were applied were linear, second degree polynomial, third degree polynomial and RBF (radial basis function; Gaussian). The above SVM models were applied on the entire sample, based on which the decision boundaries were determined, and the success of our diagnostic maps was assessed by calculating the fraction of correctly classified data. For the above analysis Matlab R2017b was used. Based on our results, we discuss which diagnostic tools offer the best classification rates and we propose next steps for their further development using other machine learning methods. The development of such diagnostics could help us test the predictions of X-Ray binary population synthesis models and obtain a better understanding of the accretion mechanism powering X-ray sources.
Language English
Subject Astronomy
Black holes
Neutron stars
Support vector machines
Αστέρες νετρονίων
Μέθοδοι μηχανικής μάθησης
Μελανές οπές
Μηχανές διανυσματικής υποστήριξης
Issue date 2019-11-29
Collection   Faculty/Department--Faculty of Sciences and Engineering--Department of Physics--Post-graduate theses
  Type of Work--Post-graduate theses
Permanent Link Bookmark and Share
Views 93

Digital Documents
No preview available

View document
Views : 6