Enhanced Higgs Boson to ${\tau }^{+}{\tau }^{-}$ Search with Deep Learning

The Higgs boson is thought to provide the interaction that imparts mass to the fundamental fermions, but while measurements at the Large Hadron Collider (LHC) are consistent with this hypothesis, current analysis techniques lack the statistical power to cross the traditional $5\sigma$ significance barrier without more data. Deep learning techniques have the potential to increase the statistical power of this analysis by automatically learning complex, high-level data representations. In this work, deep neural networks are used to detect the decay of the Higgs boson to a pair of tau leptons. A Bayesian optimization algorithm is used to tune the network architecture and training algorithm hyperparameters, resulting in a deep network of eight nonlinear processing layers that improves upon the performance of shallow classifiers even without the use of features specifically engineered by physicists for this application. The improvement in discovery significance is equivalent to an increase in the accumulated data set of 25%.

Figure 1

Distributions of low-level input variables from basic kinematic quantities in $\ell \ell +{\cancel{E}}_T$ events for simulated signal (black, solid line) and background (red, dashed line) benchmark events. Shown are the distributions of transverse momenta ($p_T$) and azimuthal angle ($\eta$) of each observed particle as well the number of hadronic jets ($N_{\text{jets}}$) and the imbalance of transverse momentum (${\cancel{E}}_T$) in the final state. Polar angle ($\varphi$) information for each observed particle is also available to the network, but is not shown, as the one-dimensional projections have little information.

Figure 2

Distributions of high-level input variables derived from the low-level variables in $\ell \ell +{\cancel{E}}_T$ events for simulated signal (black, solid line) and background (red, dashed line) benchmark events. See text for definitions.

Figure 3

Effect of network depth on discovery significance. These networks were trained with the hyperparameters optimized for the deep network.