yimuchen
commented
1 year ago Potential show-stoppers:
- Parameter initial values guessing (does an equivelant to TSpectrum [1] bump finding exist for histograms?) https://root.cern.ch/doc/master/classTSpectrum.html
- Pure function implementation for convolution [2] (Original implementation: https://github.com/UMDCMS/SiPMCalib/blob/master/SiPMCalc/src/SiPMDarkFunc.cc, does an equivalent exist for python that is still fast?)
- Many data formats formatting methods.
To disentangle this library with Yi-Mu's plotting code. Shall we consider rewriting this entire function into a python using numpy/awkward/zfit/hist/mplhep?
This also allows this analysis code to function independently of CMSSW (previously used for compile management), and deploys easily on other computation machines as well as testing.