Ranking Pulsar Plots using Coherence Measures

The methods used to detect pulsars have gone through many iterations over the years. Most recently, pulsar searching has adapted deep learning algorithms due to the many unique signatures of a pulsar detection and the amount of available data. Here, we present the first stage of development for a novel statistically coherent based approach to identifying pulsars in radio data. Based on five measures and strict cutoff values only, our method has been shown to have a recall of 0.8605, failure rate of 0.0017, and an F score of 0.8370 when tested on the HTRU Medlat Data Set. These values are competitive with those of several early neural networks built and tested on the same set.