-
Recon_bbtautau_v2.3fcb06c4a · ·
Similar to v2.1. But using flat mtautau distribution when training. Reconstruct both bb and tautau using 1 common context encoder and 2 seperate normalizing flows. Target of bb: [Px, Py, Pz, M] of truth bb - recon bb Target of tautau: [Px, Py, Pz, M] of truth tautau - (recon tautau + met)
-
Recon_bbtautau_v2.2ffa5ee8c · ·
Apart from targets specified in https://gitlab.cern.ch/cmo/nu2flow/-/tags/Recon_bbtautau_v2.1 , this tag tries to also classify [Ztautau, Htautau, continuous_Mtautau_samples] as a intermediate prediction, in order to make network better predict ttbar mass spectrum. **Result shows this doesn't make results any better.**
-
Recon_bbtautau_v2.1056c079e · ·
Updates from v2.0: target of tautau changed from "truth-recon tautau" to "truth - (recon tautau + met)". Reconstruct both bb and tautau using 1 common context encoder and 2 seperate normalizing flows. Target of bb: [Px, Py, Pz, M] of truth bb - recon bb Target of tautau: [Px, Py, Pz, M] of truth tautau - (recon tautau + met)
-
Recon_bbtautau_v2.007bf53d4 · ·
Reconstruct both bb and tautau using 1 common context encoder and 2 seperate normalizing flows. Target of bb: [Px, Py, Pz, M] of truth bb - recon bb Target of tautau: [Px, Py, Pz, M] of truth tautau - recon tautau
-
Recon_diTau_v1.4e3ba2742 · ·
Reconstruct V(truth_tautau) - V(vis_tau+MeT), where V=(Px,Py,Pz,M). And the final E is calculated using E-M relationship. Note that too many epochs can result in overtraining such that overfit M.
-
Recon_diTau_v1.3504bc65b · ·
Reconstruct diTau system four momentum using `ytautau`, `hhttbb` and `Zttxx`. Making use of jets along with more tau features as inputs.
-
Recon_diTau_v1.17fc80fdb · ·
Reconstruct total four momentum for di-Tau. Using hhttbb, Zttxx and ttbar-nonallhad.