About this video

This open access colloquium was hosted on behalf of the IOP Nuclear Physics Group by The University of Glasgow.

Machine Learning Based Electron Trigger at CLAS12
Richard Tyson, University of Glasgow
Machine learning applications in high energy physics have repeatedly demonstrated their relevance, notably for particle or channel selection. The CLAS12 Level 3 Electron trigger system was designed to efficiently select events with scattered electrons detected in the CLAS12 forward detector. However, the trigger’s low purity introduces a sizeable background which has important repercussions on data storage costs and processing times.

This talk will present a novel machine learning approach to the CLAS12 Level 3 Electron trigger based on convolutional neural networks (CNN), along with a brief introduction to machine learning and the CLAS12 detector. Preliminary tests indicate that the CNN trigger could lead to a ~70% data reduction rate compared to the current trigger whilst also maintaining above 99.5% efficiency.

AI for data quality monitoring
Thomas Britton, Jefferson Lab
Data quality monitoring is a critical component to all experiments and directly impacts the overall quality of an experiment's physics results. Traditionally, this is done through a low level alarm system (e.g. detector voltages), leaving higher level monitoring to human crews. Artificial Intelligence based technologies are beginning to find their way into scientific applications, but this comes with difficulties; often relying on the acquisition of new skill sets, either through education or acquisition, in data science. This talk will describe Hydra, an AI based monitoring system deployed at GlueX. It will also describe how the use of "off-the-shelf" technologies can drastically reduce the time to deployment of such systems as well as discuss what sociological hurdles must be overcome to successfully deploy such a system.

AI for experiment control and calibration
Naomi Jarvis, Carnegie Mellon University
The artificial intelligence (AI) for experiment control and calibration project at Jefferson Lab, VA, aims to develop and deploy an artificial intelligence system to control and calibrate detector systems. Drift chambers are an ideal subject for this: they are used widely for tracking and particle identification, but their calibration can be time-consuming as each chamber's performance is dependent on environmental factors such as pressure, temperature and the flux of incident particles.

The AI system under development will monitor the environmental quantities and suggest a new high voltage setting for the detector at the time of measurement, to maintain consistent gain and resolution throughout the weeks of data-taking. It will also predict the values necessary to calibrate the data.

This talk will describe development of the AI system for the GlueX Central Drift Chamber.