Neural conditional reweighting

There is a growing use of neural network classifiers as unbinned, high-dimensional (and variable-dimensional) reweighting functions. To date, the focus has been on marginal reweighting, where subset features are used for while all other integrated over. some situations, though, it preferable to condition auxiliary instead marginalizing over them. In this paper, we introduce conditional which extends case. This approach particularly relevant in high-energy physics experiments detector effects conditioned particle-level truth information. We leverage custom loss function that not only allows us achieve through single training procedure, but also yields sensible interpolation even presence phase space holes. As specific example, apply energy response jets, could be improve modeling objects parametrized fast simulation packages.