search

ML4GW / aframe

Detecting binary black hole mergers in LIGO with neural networks
MIT License
18 stars 17 forks source link

Investigate SV dip for production curves/apply vetoes #440

Closed alecgunny closed 9 months ago

alecgunny commented 10 months ago

As it stands, the two re-seeded production training runs have noticeable dips in SV at the lowest FARs, indicating possibly anomalously loud background events.

  1. Is this actually the case? Do their foreground distributions match those from earlier runs?
  2. Are these events nearly the same for both models?
  3. What happens when we apply vetoes?
alecgunny commented 10 months ago

@EthanMarx I can tackle (1) if you can tackle (2) and (3)

EthanMarx commented 10 months ago

Sounds good!

EthanMarx commented 9 months ago

It was discovered that this dip was due to a known event contaminating the background in a timeslide.