Random Forest Improves Gravitational-Wave Candidate Detection
Researchers present a Random Forest classifier applied to O3a and O3b LIGO-Virgo data to improve template-based gravitational-wave candidate ranking. Evaluated on double-coincidence events from the MBTA pipeline, the classifier yields a modest but consistent increase in detections at low false positive rates and computes p_astro for each event. Using this statistic, the team recovers catalogued events and identifies one new subthreshold candidate (IFAR=0.05, gps 1240423628).
Key Points
- 1Apply Random Forest classifier to O3a/O3b double-coincidence events, using MBTA pipeline outputs.
- 2Achieve modest but consistent detection increase at low false positive rates versus standard re-weighted SNR.
- 3Compute p_astro per event and identify a new subthreshold candidate (IFAR=0.05), aiding follow-up.
Scoring Rationale
Moderately novel application of supervised ML to GW candidate ranking with clear practitioner relevance and a concrete new subthreshold candidate. Scored for moderate novelty and strong relevance to ML-in-physics; credibility limited by arXiv preprint status but boosted by timely, reproducible evaluation on O3a/O3b data.
Sources
Public references used for this report.
Practice interview problems based on real data
1,625 SQL & Python problems across 15 industry datasets — the exact type of data you work with.
Try 250 free problems
