ff.flare_table comes out with fewer flares than the tutorial indicates

[ekaterinailin]

Hi @afeinstein20 !

I tried to use the FitFlares feature, but got stuck at the point where I wanted to show the table with the fitted flares.

This is the code I ran (trying to reproduce the Fitting Flares tutorial):

import stella

# Get the light curve

from lightkurve.search import search_lightcurvefile

lk = search_lightcurvefile(target='tic62124646', mission='TESS')
lk = lk.download().PDCSAP_FLUX

# get the model

cnn = stella.ConvNN(output_dir=OUT_DIR)

MODEL_DIR = 'run01/'
MODEL = [os.path.join(MODEL_DIR,i) for i in
          os.listdir(MODEL_DIR) if i.endswith('.h5')][0]

# predict flares from model

cnn.predict(modelname=MODEL,
            times=lk.time,
            fluxes=lk.flux,
            errs=lk.flux_err)

# fit flares

ff = stella.FitFlares(id=[lk.targetid],
                      time=[lk.time],
                      flux=[lk.flux],
                      flux_err=[lk.flux_err],
                      predictions=[cnn.predictions[0]])

# get flare table

ff.identify_flare_peaks(threshold=0.5)

ff.flare_table

The output is a single flares

Target_ID	tpeak	amp	ed_s	rise	fall	prob
62124646.0	1662.6966771763764	1.0031577638536382	2.5355761288160976	0.0028550217295811296	0.008296178315620328	0.976371169090271

which is not what I expected. Based on the cnn.predictions I expected a number of flares similar to what your tutorial shows (16?).

I tried several models from the quickstart collection, but they either return 1 or zero flares in the ff.flare_table.

Do you have an idea what I am doing wrong here?

Cheers :) Ekaterina

[afeinstein20]

Hi @ekaterinailin ! Weird. Which version of stella are you using? If you're using the version that was pip install stella then I would believe there would be differences between that and the version now on GitHub.

I will be updating the PyPI version once the referee report is good to go and the associated AJ paper is accepted!

[ekaterinailin]

I think I tried both the pip version and the cloned repo. I now reproduced the same result with the most recent master branch version.

[afeinstein20]

Okay great! Yeah I think there was a flub in me working in different branches, but now there's consistency between them. Please close the issue if you're satisfied with the fix!

[ekaterinailin]

I got a fresh installation from the master branch now, but unfortunately to no avail. I get a bunch of outputs tothe prompt, not sure if related to the actual issue, but perhaps worth posting anyways. Some I expected from the tutorial, others not:

/home/ekaterina/Documents/000_envs/stella/lib/python3.6/site-packages/lightkurve/lightcurvefile.py:47: LightkurveWarning: `LightCurveFile.header` is deprecated, please use `LightCurveFile.get_header()` instead.
  LightkurveWarning)

WARNING: No stella.DataSet object passed in.
Can only use stella.ConvNN.predict().

100%|██████████| 1/1 [00:01<00:00,  1.13s/it]
Finding & Fitting Flares:   0%|          | 0/1 [00:00<?, ?it/s]/home/ekaterina/Documents/000_envs/stella/lib/python3.6/site-packages/stella-0.1.0rc1-py3.6.egg/stella/mark_flares.py:92: VisibleDeprecationWarning: Creating an ndarray from ragged nested sequences (which is a list-or-tuple of lists-or-tuples-or ndarrays with different lengths or shapes) is deprecated. If you meant to do this, you must specify 'dtype=object' when creating the ndarray
  return np.array(results)
/home/ekaterina/Documents/000_envs/stella/lib/python3.6/site-packages/numpy/lib/nanfunctions.py:1665: RuntimeWarning: Degrees of freedom <= 0 for slice.
  keepdims=keepdims)
/home/ekaterina/Documents/000_envs/stella/lib/python3.6/site-packages/numpy/lib/nanfunctions.py:1114: RuntimeWarning: All-NaN slice encountered
  overwrite_input=overwrite_input)
Finding & Fitting Flares: 100%|██████████| 1/1 [00:00<00:00,  5.22it/s]

[danirevi5]

Hi,

I am having the same problem, i find fewer flares than in the tutorial. I have tried both, the pip and the github installation and they all give the same result.

In other cases, although the star I am working with has many obvious flares, it only detect two small ones but not the obvious ones.

Has there been any change that could be causing this?

Thank you!

[aftab234]

I get 16 flares as in the tutorial if I don't normalize the light curve and use the remove_nans() function along with the s004 pre-trained model. But the equivalent duration is in seconds now in the most recent version of stella, instead of minutes.

Edit: Not the same 16 flares. There are duplicates! 67 flares, again with many duplicates, if I normalize the light curve.