search

afeinstein20 / stella

For characterizing flares with convolutional neural networks
https://adina.feinste.in/stella
MIT License
26 stars 18 forks source link

Duplicates #18

Open benjaminpope opened 3 years ago

benjaminpope commented 3 years ago

Hi Adina,

Reprocessing all our data for the paper using the latest version of stella (more on that when finished!)

Noticed a funny bug: take (say) the 120s data for 2MASS J09481615+5114518 in Sector 21. We get the avg_preds from the ensemble prediction over the new models (seems fine) and obtain a flare table with

ff = stella.FitFlares(id=tics, time=time, flux=flux, predictions=avg_preds, flux_err=errs) ff.identify_flare_peaks(threshold=0.6)

we get many duplicates in the flare table.

These are exact duplicates so you can just filter them with unique(flare_table) but concerned there is a deeper bug?

Cheers,

Ben

benjaminpope commented 3 years ago

Also finding quite a few flares missed in CR Dra in fit_flares vs spikes with high CNN probability that look good by eye...