Higgs Boson Detection from Monte Carlo Simulation Tensors using Wide and Deep Networks

Higgs Bosons are the fundamental particles with 0 spin that couple to mass and are produced by quantum excitation of the Higgs Field. They have even parity and have no electric or colour charge. It is very unstable in nature and quickly decays into other particles. Monte Carlo simulation is a probabilistic technique to estimate different outcomes in a process that are difficult to predict because of random variables. The Monte Carlo simulation method can be used to generate a probabilistic dataset containing 211ow-level features and 7 high-level features to detect Higgs Boson. Deep learning, on the other hand, is very useful to create a decision model based on a large amount of data with fast prediction capabilities. But training a deep learning model comes with requiring large computational resources. Other related works which have been conducted on this domain take a large number (50 to several thousand) of iterations to converge to a stable loss. The presented paper aims to reduce the number of iterations and resource consumption to train the dataset by using a Wide and Deep Network based learning model. The presented model converges to an AUC of 0.89 within 12 iterations hitting an early stopping condition. This shows that the presented model is more efficient than other available models.