Neural Network Team

Title: A Comparative Study of Neural Networks and Logistic Regression for High Energy Physics

Team Members: Joel Gonzalez-Santiago, Nigel Pugh, Thomas Johnson III

Mentor: Jerome E. Mitchell

Abstract:  Neural networks are programs that run based on machine learning algorithms and resources to mirror the function of the brain in its roughest capacity. Neural networks are used primarily for the management and manipulation of large quantities data to form classification, more efficient searches, and prediction of the data. Neural networks exist as part of the larger field of machine learning that exists. Linear regression in turn serves as the statistics based solution to the classification issue, an alternative to neural networks that are also a form of machine learning. The focus of this research was to observe whether neural networks or linear regression models are more effective for classification of a supersymmetry dataset. The supersymmetry dataset is made up of the results gathered particle collision events within a particle accelerator. Supersymmetry itself is a theory within particle physics that suggests the particles that are absent in the standard model are symmetric, or balancing, counterparts to the particles that have been already discovered.

Keywords: Neural Networks, High Energy Physics, Logistic Regression

Extended Abstract: PDF DOC