Aim: to classify the reliability of Δν measurements using neural networks
Inputs required:
- Background corrected, long cadence power spectra, individual files should end in '.psd.bgcorr'
- Table containing pipeline results of estimated ν_{max} and Δν, in coloumns named "numax(gaus)" and "dnu", respectively. Column named "file" should correspond to the names of the PSD files, but ending on '.psd' (not '.psd.bgcorr')
Outputs:
- histogram_NAMEDATA.png
- results_NAMEDATA.png
- NAMEDATA_ALL_N1_GOODDNU_N2_FP_N3
Where NAMEDATA is the name of the folder containing the power spectra, N1 is the number of stars (or PSD) with dnu and numax results from the pipeline, N2 is the number of dnu with probabilities exceeding the threshold imposed, and N3 is the number of false positives identified because their numax/dnu values stray too much from the empyrical relation.
Files 1 and 2 are useful graphical representations of the results. File 3 is the table with the calculated probabilities by the neural network.
===============
Author: Claudia Reyes
Dates: 2020-2024
===============
Published: Monthly Notices of the Royal Astronomical Society, Volume 511, Issue 4, April 2022, Pages 5578–5596
Title: Vetting asteroseismic Δν measurements using neural networks
Authors: Claudia Reyes, Dennis Stello, Marc Hon, Joel C Zinn
DOI: https://doi.org/10.1093/mnras/stac445
===============
Preferred python version:
- 3.8
Followind modules are used, and normally part of basic python environment:
- numpy
- matplotlib
- pandas
- os
- re
- math
- scipy
Special modules cv2 and sklearn likely need to be installed with the commands:
> pip install scikit-learn
> pip install tensorflow==2.13.1
> pip install opencv-python
If opencv-python fails to install, try first:
> pip install --upgrade pip setuptools wheel
Download the project folder to your machine and:
- Open Dnu_classifier.py using a text editor
- Locate the lines between "INPUT" and "INPUT END"
- Replace the paths with the corresponding path of the user machine
- If needed, replace the value of "threshold" with a value between 0 and 1.
- Save changes to Dnu_classifier.py
- run the classifier from terminal using the command "python Dnu_classifier.py" this may take between 3 and 20+ minutes, depending on the local machine and the amount of data.