MSPAC (Machine Sequenced PulsAr Candidates)  

Neutron stars are the super-dense remnants of dead stars: too massive to form a white dwarf but not massive enough to form black holes. A typical neutron star is about 1.5 times the mass of our sun, but its radius is only about 10-15 kilometers. Pulsars are rapidly spinning neutron stars that emit radio waves from their magnetic poles. We have discovered more than 3,000 pulsars through manual search methods. Past researchers at Kenyon College have created PACMANN (PulsAr-Classifier Machine-Learning Algorithm with Neural Networks), which is a machine learning algorithm that can sift through pulsar candidates and identify the most promising ones. However, the accuracy of any machine learning algorithm relies heavily on the robustness of the training data set. My research works to fill in a vital missing piece of PACMANN’s training data set: a lack of confirmed pulsars on the scale required to train a machine learning algorithm. One solution to this deficit in the training data set is to simulate real pulsars. We developed such an algorithm, named MSPAC (Machine-Sequenced PulsAr Candidates), which simulates a pulsar candidate population from distributions of real pulsars. Our goal with these machine-sequenced pulsars is to create a pulsar population indistinguishable from real pulsar populations and use these populations to train PACMANN.