Sep 5 – 10, 2021
Online
Europe/Lisbon timezone

A Machine Learning Algorithm for Triggering the Project 8 Neutrino Mass Experiment

Sep 7, 2021, 7:04 PM
1m
Online

Online

Poster Neutrino physics Poster Session II

Speaker

Andrew Ziegler (Penn State)

Description

Project 8 is a next-generation neutrino mass experiment that uses Cyclotron Radiation Emission Spectroscopy (CRES) to measure the neutrino mass. CRES is a novel technique for $\beta$-decay spectroscopy that measures the frequency of the cyclotron radiation produced by energetic electrons trapped in a magnetic field. The cyclotron frequency can be directly converted into the energy spectrum, which yields the neutrino mass through measurement of the spectrum endpoint. The next phase of Project 8 seeks to measure the energy spectrum of molecular Tritium $\beta$- decay in an O(10 cm$^3$) free space volume, using a multi-channel phased array of antennas. The low signal power (< 1fW) and multi-channel reconstruction place stringent constraints on the triggering and online signal processing algorithms. I present progress on a machine learning triggering algorithm that employs a Deep Convolutional Neural Network (DCNN) to detect the presence of cyclotron radiation signals buried in noise. The network achieves greater than 85% classification accuracy on simulated electron signals, outperforming conventional methods, and rivaling the performance of an optimal matched filter while requiring significantly fewer computational resources for online operation.

Primary author

Andrew Ziegler (Penn State)

Presentation materials