Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Repeating versus Nonrepeating Fast Radio Bursts: A Deep Learning Approach to Morphological Characterization

Bikash Kharel, Emmanuel Fonseca, Charanjot Brar, Afrokk Khan, Lluis Mas-Ribas, Swarali Shivraj Patil, Paul Scholz, Seth Robert Siegel, David C. Stenning

TL;DR

This paper tackles distinguishing repeating from non-repeating FRBs using purely morphological information extracted from CHIME/FRB Catalog 2 dynamic spectra. It adopts a transfer-learning approach with ConvNeXt, treating dedispersed dynamic spectra as images and fine-tuning on total-intensity data to capture morphology. Key contributions include demonstrating persistent morphological differences between the two sub-classes in Catalog 2, using image-based features rather than model-based fits, and achieving high metrics with reduced training time. The method enables rapid inference for new events and supports morphology-driven FRB modeling and follow-up observations.

Abstract

We present a deep learning approach to classify fast radio bursts (FRBs) based purely on morphology as encoded on recorded dynamic spectrum from CHIME/FRB Catalog 2. We implemented transfer learning with a pretrained ConvNext architecture, exploiting its powerful feature extraction ability. ConvNext was adapted to classify dedispersed dynamic spectra (which we treat as images) of the FRBs into one of the two sub-classes, i.e., repeater and non-repeater, based on their various temporal and spectral properties and relation between the sub-pulse structures. Additionally, we also used mathematical model representation of the total intensity data to interpret the deep learning model. Upon fine-tuning the pretrained ConvNext on the FRB spectrograms, we were able to achieve high classification metrics while substantially reducing training time and computing power as compared to training a deep learning model from scratch with random weights and biases without any feature extraction ability. Importantly, our results suggest that the morphological differences between CHIME repeating and non-repeating events persist in Catalog 2 and the deep learning model leveraged these differences for classification. The fine-tuned deep learning model can be used for inference, which enables us to predict whether an FRB's morphology resembles that of repeaters or non-repeaters. Such inferences may become increasingly significant when trained on larger data sets that will exist in the near future.

Repeating versus Nonrepeating Fast Radio Bursts: A Deep Learning Approach to Morphological Characterization

TL;DR

This paper tackles distinguishing repeating from non-repeating FRBs using purely morphological information extracted from CHIME/FRB Catalog 2 dynamic spectra. It adopts a transfer-learning approach with ConvNeXt, treating dedispersed dynamic spectra as images and fine-tuning on total-intensity data to capture morphology. Key contributions include demonstrating persistent morphological differences between the two sub-classes in Catalog 2, using image-based features rather than model-based fits, and achieving high metrics with reduced training time. The method enables rapid inference for new events and supports morphology-driven FRB modeling and follow-up observations.

Abstract

We present a deep learning approach to classify fast radio bursts (FRBs) based purely on morphology as encoded on recorded dynamic spectrum from CHIME/FRB Catalog 2. We implemented transfer learning with a pretrained ConvNext architecture, exploiting its powerful feature extraction ability. ConvNext was adapted to classify dedispersed dynamic spectra (which we treat as images) of the FRBs into one of the two sub-classes, i.e., repeater and non-repeater, based on their various temporal and spectral properties and relation between the sub-pulse structures. Additionally, we also used mathematical model representation of the total intensity data to interpret the deep learning model. Upon fine-tuning the pretrained ConvNext on the FRB spectrograms, we were able to achieve high classification metrics while substantially reducing training time and computing power as compared to training a deep learning model from scratch with random weights and biases without any feature extraction ability. Importantly, our results suggest that the morphological differences between CHIME repeating and non-repeating events persist in Catalog 2 and the deep learning model leveraged these differences for classification. The fine-tuned deep learning model can be used for inference, which enables us to predict whether an FRB's morphology resembles that of repeaters or non-repeaters. Such inferences may become increasingly significant when trained on larger data sets that will exist in the near future.

Paper Structure

This paper contains 6 sections, 3 figures.

Table of Contents

  1. Introduction
  2. Images from CHIME/FRB Data
  3. A CNN Model for CHIME/FRB Images
  4. Fine Tuning on Total Intensity Data
  5. Data Preparation and Preprocessing
  6. Analysis and Results

Figures (3)

  • Figure 1: Typical dynamic spectra of FRBs showing different burst morphology in different subclasses. (Left) A repeating FRB with narrow bandwidth and broader pulse width. (Right) An apparently non-repeating broad-band FRB with narrow pulse width. Any other bursts from the source haven't been recorded to date.
  • Figure 2: Corner plot for different morphological parameters in CHIME/FRB Catalog 2. Blue denotes repeating FRBs and gray denotes as yet non-repeating FRBs.
  • Figure 3: Distribution of latent-space features of burst morphologies for three CHIME/FRB repeating sources with the largest number of events, extracted using an autoencoder. The latent features were reduced to two dimensions with t-SNE.