Date of Graduation
Statler College of Engineering and Mineral Resources
Lane Department of Computer Science and Electrical Engineering
Automated single-pulse search approaches are necessary as ever-increasing amount of observed data makes the manual inspection impractical. Detecting radio pulsars using single-pulse searches, however, is a challenging problem for machine learning because pul- sar signals often vary significantly in brightness, width, and shape and are only detected in a small fraction of observed data.
The research work presented in this dissertation is focused on development of ma- chine learning algorithms and approaches for single-pulse searches in the time domain. Specifically, (1) We developed a two-stage single-pulse search approach, named Single- Pulse Event Group IDentification (SPEGID), which automatically identifies and clas- sifies pulsars in radio pulsar search data. SPEGID first identifies pulse candidates as trial single-pulse event groups and then extracts features from the candidates and trains classifiers using supervised machine learning. SPEGID also addressed the challenges in- troduced by the current data processing techniques and successfully identified bright and dim candidates as well as other types of challenging pulsar candidates. (2) To address the lack of training data in the early stages of pulsar surveys, we explored the cross-surveys prediction. Our results showed that using instance-based and parameter-based transfer learning methods improved the performance of pulsar classification across surveys. (3) We developed a hybrid recommender system aimed to detect rare pulsar signals that are often missed by supervised learning. The proposed recommender system uses a target rare case to state users’ requirements and ranks the candidates using a similarity func- tion which is calculated as a weighted sum of individual feature similarities. Our hybrid recommender system successfully detects both low signal-to-noise ratio (S/N) pulsars and Fast Radio Bursts (FRBs).
The approaches proposed in this dissertation were used to analyze data from the Green Bank Telescope 350 MHz drift (GBTDrift) pulsar survey and the Arecibo 327 MHz (AO327) drift pulsar survey and discovered eight pulsars that were overlooked in previous analysis done with existing methods.
Pang, DI, "Identification and Classification of Radio Pulsar Signals Using Machine Learning" (2021). Graduate Theses, Dissertations, and Problem Reports. 10280.