Separating repeating fast radio bursts using the minimum spanning tree as an unsupervised methodology

TL;DR

This paper introduces an unsupervised graph theory approach using Minimum Spanning Trees to classify fast radio bursts as repeaters or non-repeaters, offering a new tool for astrophysical data analysis.

Contribution

The study applies MST methodology to classify FRBs, demonstrating its effectiveness compared to machine learning and proposing potential repeater candidates.

Findings

MST-based method effectively separates repeaters from non-repeaters.

The approach outperforms some existing machine learning classifiers.

Potential new repeater candidates are identified.

Abstract

Fast radio bursts (FRBs) represent one of the most intriguing phenomena in modern astrophysics. However, their classification into repeaters and non-repeaters is challenging. Here, we present the application of the graph theory Minimum Spanning Tree (MST) methodology as an unsupervised classifier of repeaters and non-repeaters FRBs. By constructing MSTs based on various combinations of variables, we identify those that lead to MSTs that exhibit a localized high density of repeaters at each side of the node with the largest betweenness centrality. Comparing the separation power of this methodology against known machine learning methods, and with the random expectation results, we assess the efficiency of the MST-based approach to unravel the physical implications behind the graph pattern. We finally propose a list of potential repeater candidates derived from the analysis using the MST.