Chat Chris Berman 15 Aug 2024

Will try to predict the mA distributions with simple PCA and check if it beats the dumb pt binning. Can check this by binning finer than the binning used to make the dumb templates and see if there is a better description of the mA within the bins. If that works implement the PCA with an autoencoder (eg: use linear activation functions) this should give the same result. If that works can then move to non-linear activiations and improve further.