Session 3F: Applications of Artificial Intelligence in Astrophysics
Session Number
Session 3F: 1st Presentation
Advisor(s)
Dr. Brian Nord, Fermilab
Location
Room A113
Start Date
26-4-2018 12:40 PM
End Date
26-4-2018 1:25 PM
Abstract
Deep sky surveys have the potential to reveal more information about dark matter and dark energy. While sorting through the abundance of data from these surveys can be cumbersome, machine learning algorithms present a more efficient method of doing so. In this project, we aim to develop a computerized method of detecting phenomena, specifically Einstein Rings, in sky surveys.
We start with exploring gravitational lensing to understand the physical explanation of Einstein rings. Einstein Rings are characterized by their near-perfect circular shape, so we created a python program that outputs a random assortment of circles and polygons as image files. We then wrote a program that can distinguish between a circle and a polygon using the Hough Transform, and tested it on the simulated images from the first program. We performed some basic statistical diagnostics – calculating the specificity and sensitivity -- to get an idea of how well the circle detection program worked. We are currently modifying the program accordingly to improve its ability to correctly identify whether circles are present in an image as well as find the centers and radii of any detected circles. The next step is to test the program on images from real sky surveys.
Session 3F: Applications of Artificial Intelligence in Astrophysics
Room A113
Deep sky surveys have the potential to reveal more information about dark matter and dark energy. While sorting through the abundance of data from these surveys can be cumbersome, machine learning algorithms present a more efficient method of doing so. In this project, we aim to develop a computerized method of detecting phenomena, specifically Einstein Rings, in sky surveys.
We start with exploring gravitational lensing to understand the physical explanation of Einstein rings. Einstein Rings are characterized by their near-perfect circular shape, so we created a python program that outputs a random assortment of circles and polygons as image files. We then wrote a program that can distinguish between a circle and a polygon using the Hough Transform, and tested it on the simulated images from the first program. We performed some basic statistical diagnostics – calculating the specificity and sensitivity -- to get an idea of how well the circle detection program worked. We are currently modifying the program accordingly to improve its ability to correctly identify whether circles are present in an image as well as find the centers and radii of any detected circles. The next step is to test the program on images from real sky surveys.