Identification of Boosted Higgs Bosons Decaying Into $b\bar{b}$ With Neural Networks and Variable Radius Subjets in ATLAS

Boosted Higgs bosons decaying via the dominant $H\to b\bar{b}$ mode are an essential ingredient to a number of LHC physics signatures. ATLAS identifies these hadronic Higgs bosons by reconstructing a large-$R$ ($R=1.0$) jet, associating variable-radius subjets to the large-$R$ jet, and then assigning a flavor discriminant to each of the subjets using standard flavor tagging. This work introduces an algorithm to select boosted Higgs bosons that combines flavor discriminants from up to three subjets using a feed-forward neural network. A comparison of several algorithms that combine individual subjet discriminants show that the neural network-based algorithm is the most powerful, assessed on its ability to reject both boosted top quark jets and jets arising from multijet processes.