Searching for New Physics through Transparent Machine Learning
Abstract: Significant advances have been made in developing powerful machine learning algorithms such as boosted decision trees and, more recently, convolutional neural networks. The transparent use of these machine learning methods in the search for new physics involving tops and missing transverse energy with ATLAS is presented. Future improvements to the sensitivity to soft tops (pT~200 GeV) plus missing transverse energy are explored by using boosted decision trees and convolutional neural networks. Additionally, improvements to Phase-I and Phase-II calorimeter trigger efficiency for soft top production using convolutional neutral network are discussed.