Statistical Inference for a Particle Simulator

• Activity Number: 373
• Type: Topic Contributed
• Date/Time: Wednesday, August 11, 2004 : 2:00 PM to 3:50 PM
• Sponsor: Section on Bayesian Statistical Science
• Author(s): Herbert Lee*+ and Bruno Sanso and Weining Zhou and David Higdon
• Companies: University of California, Santa Cruz and University of California, Santa Cruz and University of California, Santa Cruz and Los Alamos National Laboratory
• Keywords: inverse problem ; proton accelerator ; Gaussian process

Abstract

A beam of protons is produced by a linear charged particle accelerator, then focused through the use of successive quadrupoles. The initial state of the beam is unknown, in terms of particle position and momentum. Wire scans are used to collect data on the current state of the beam as it passes through and beyond the focusing region, and the goal is to infer the initial state from the wire trace data. This setup is that of a classic inverse problem, in which a computer simulator is used to link an initial state configuration to observable values (wire traces), and then inference is performed for the distribution of the initial state. We model the initial distribution (position and momentum) as two bivariate Gaussian processes, one for each of the x and y directions. The process convolution approach improves the computational efficiency.