Open Access. Powered by Scholars. Published by Universities.®
Physical Sciences and Mathematics Commons™
Open Access. Powered by Scholars. Published by Universities.®
- Discipline
Articles 1 - 3 of 3
Full-Text Articles in Physical Sciences and Mathematics
Optimizing Convolutional Neural Networks For Transient Detection In Optical Astronomy With Augmented Datasets, Wendy Mendoza
Optimizing Convolutional Neural Networks For Transient Detection In Optical Astronomy With Augmented Datasets, Wendy Mendoza
Theses and Dissertations
We present a technique for optical transient detection using artificial neural networks, particularly a Convolutional Neural Network (CNN), a deep learning algorithm. This method analyzes images of the same area of the sky captured by several telescopes, with one image serving as a reference for a probable transient’s epoch and the other as an image from a previous epoch. We train the CNN on simulated sources and test it on actual image data samples using data from the Dr. Cristina V. Torres Memorial Astronomical Observatory and Sloan Digital Sky Survey. This autonomous detection method replaces the standard procedure, which involves …
Searching For Low Frequency Fast Radio Bursts With Vlite, Suryarao Bethapudi
Searching For Low Frequency Fast Radio Bursts With Vlite, Suryarao Bethapudi
Theses and Dissertations
The VLITE (VLA Low Band Ionosphere and Transient Experiment; http://vlite.nrao.edu) program performs commensal observations using 16 antennas of the Very Large Array radio telescope from 320-384 MHz. The VLITE-Fast program searches for short time-scale (<100ms) transients, such as Fast Radio Bursts (FRBs), in real time and triggers recording of baseband voltages for offline imaging. Searches are made possible by a 12 node cluster, each housing GPUs for digital signal processing. A real-time Message Passing Interface (MPI)-based co-adder incoherently sums the data streams from all the antennas to boost the signal-to-noise. To undo the dispersion effects of signal propagation through the ionized interstellar medium, the co-added stream is de-dispersed and matched-filtered to search for transients. This operation is completely performed on GPUs by the software package Heimdall . A selection logic is applied to the candidates and interesting candidates with their corresponding data are processed and packaged in a binary file along with a diagnostic plot. Furthermore, a Machine Learning classification is applied on the reduced data product and, based on its decision, baseband voltages are recorded. Reduced data products collected over 126 days of on-sky operation form the VLITE-Fast Pathfinder Survey (VFPS). This pipeline has triggered on single pulses from 7 known radio pulsars. Lastly, the pipeline capabilities are tested against pure random noise and simulated injected signals.
Pipeline For Variable Star Detection And Eclipsing Binary Characterization, Moises Castillo
Pipeline For Variable Star Detection And Eclipsing Binary Characterization, Moises Castillo
Theses and Dissertations
Stars have been observed and recorded since ancient times. Practices of documenting brightness led to observations of variability. Optical CCD observations of eclipsing binary stars were made with instruments at UTRGV Dr. Cristina Valeria Torres Memorial Astronomical Observatory. There are two main goals for this project. The first goal is to create a pipeline written in python (lightcurator) that creates a framework for detecting variable stars. The pipeline starts by creating a list of ccd frames written in FITS format of an eclipsing binary star observation. These object frames are expected to be already reduced, but lightcurator provides tools to …