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 = 9.0rfi_threshold = 5.0high_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.
