Feature Study And Prediction of Repeater Candidates

[Thaza-Kun]

Description

Dimensional reduction algorithms such as T-SNE (doi:10.48550/arXiv.2210.02471) and UMAP (doi:10.1093/mnras/stab2994) shows clusters of FRBs in their own respective spaces suggesting there is hidden similarity between them. It is possible that different clusters represents different FRB emission types. It is also not unjustified to assume a single emission type covers multiple clusters. This assumption allows all of them to share one emmision mechanism but does not disqualify multiple emission mechanism.

Prerequisite

• [ ] Dimensional reduction algorithm
• [x] A dataset with a significant population
• [ ] A selection of parameters
• [ ] Collect periodicity of known repeaters

Proposed Methodology

• Xiang-Han et al (2020) doi:10.1093/mnras/staa3351 uses the Mann–WhitneyWilcoxon (M–W–W) test to test whether repeating and non-repeating FRBs may originate from different mechanism using pulse width and radio luminosity. Anderson-Darling test is also used to test whether these parameters follow a Gaussian distribution.
• A regression model might be used with Leave-One-Out K to predict values of known properties within cluster to test the feasability of the prediction. Next, we can predict the periodicity of repeater candidates.

[Thaza-Kun]

Important Parameters

Using Chen et al 2021 and Jia-Wei 2022 as reference, we can infer that these four are significant parameters:

1. highest frequency
   • which is available from telescopic data. CHIME/FRB data (~500) is available using Hashimoto #3. We need to dig other telescope's data for the remaining 200 - 300 subbursts.
2. peak frequency
   • which is the frequency which it is the brightest
3. frequency width
   • which is the difference between the highest and the lowest frequency
4. brightness temperature
   • which can be derived according to Jia-Wei 2022.

[Thaza-Kun]

22 partially satisfy this trajectory (but not using machine learning)