Abstract
We compared convolutional neural networks to the classical boosted decision trees for the separation of atmospheric particle showers generated by gamma rays from the particle-induced background. We conduct the comparison of the two techniques applied to simulated observation data from the Cherenkov Telescope Array. We then looked at the Receiver Operating Characteristics (ROC) curves produced by the two approaches and discuss the similarities and differences between both. We found that neural networks overperformed classical techniques under specific conditions.
Highlights
Machine learning made spectacular advances during the last few years
In this paper we focus on the signal extraction and we evaluate the performance of convolutional neural networks (CNNs) compared to Boosted Decision Trees (BDTs) [3] which are commonly used for this task and in particular in the EventDisplay analysis package [4]
Even though results are presented in the form of Receiver Operating Characteristics (ROC) curves below, in reality the ratio of signal/background events is in the order of
Summary
Machine learning made spectacular advances during the last few years. Deep convolutional neural networks (CNNs) emerged as a very powerful technique thanks to advances in algorithms, data availability and overall computational power. In this paper we focus on the signal extraction and we evaluate the performance of CNNs compared to Boosted Decision Trees (BDTs) [3] which are commonly used for this task and in particular in the EventDisplay analysis package [4]. We took the Event parameter output of the EventDisplay analysis of simulated events from a Monte-Carlo (MC) production of CTA.
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
