juppy44/frb-rfi-classifier-chime-injections

FRB vs RFI Classifier on CHIME/FRB Injections

This repository contains artifacts produced by a Google Colab notebook that trains multiple tabular classifiers on the CHIME/FRB Catalog 1 injections dataset.

Task

Binary classification:

• 1 = detected astrophysical-like event
• 0 = RFI

Label logic

Labels follow the public CHIME injections tutorial logic:

• snr_threshold = 9.0
• rfi_threshold = 5.0
• high_snr_override = 100.0

Models trained

• Logistic Regression
• Random Forest
• XGBoost
• Small PyTorch MLP

Best model

Random Forest

Notes

• This is a research prototype, not a production real-time FRB pipeline.
• The public injections release is tabular metadata for injected events. It is a good fit for structured ML baselines.