Support this project and keep always updated by following on github.

deepIRF

deepIRF - A deep learning approach for reverse-broadening the instrumental response (IRF) in lifetime spectra.

`preliminary results`

Is it generally possible to correct for the timing uncertainties inherently produced by the photomultiplier tubes?

1 Mio. pulse-pairs recorded from an isotope 60-Co have been stored along with the obtained timing differences (target/label value), i.e. the PMT uncertainty between them
70 % of the pulse-pairs have been used for training the model
the other 30 % of the pulse-pairs have been used to test the trained model
the results as shown in the figure below indicate that in principle deep learning can significantly reduce the uncertainty originating from the PMTs
blue curve = initial "unseen" data (30 %), red curve = deepIRF-corrected initial "unseen" data (30 %)