MLForSplittingFunctions
Can we improve PDF sampling in the jets (de)clustering with Maching Learning ? Maybe we can use explicit MC to get splitting functions ?
Follow-up
- Test by seeing that we reproduce the 1D marginals
- Test the 1D marginals conditioned on comb pt
- Test one of zA or thetaA marginal conditions on the other
- check the compression (eg: size of data stored)
- Think what we want is some kind of generative model
- eg: auto encoder parameterized with pT, pT + rand -> {z, θ, phi, m, m}