In BinnedLogLikelihood we set the data with the bins defined by analysis space:
https://github.com/JelleAalbers/blueice/blob/f12c2adffc0698e874a5fa25cbb7436fc5c58d54/blueice/likelihood.py#L488-L490
Therefore, the data is automatically chopped. However in UnbinnedLogLikelihood there is no such an operation. As the pdfs are interpolated & extrapolated, the program doesn't know which data points are out of bound, and therefore perform a lot of redundant calculations for the cost function which slow down the call a lot.
In
BinnedLogLikelihood
we set the data with the bins defined by analysis space: https://github.com/JelleAalbers/blueice/blob/f12c2adffc0698e874a5fa25cbb7436fc5c58d54/blueice/likelihood.py#L488-L490 Therefore, the data is automatically chopped. However inUnbinnedLogLikelihood
there is no such an operation. As the pdfs are interpolated & extrapolated, the program doesn't know which data points are out of bound, and therefore perform a lot of redundant calculations for the cost function which slow down the call a lot.