Machine Learning methods for Directional Recoil Detection reconstruction at CYGNUS-Oz

25 Feb 2026, 09:40
25m
Integrated Research Center (Kobe University)

Integrated Research Center

Kobe University

7 Chome-1-48 Minatojima Minamimachi, Chuo Ward, Kobe, Hyogo 650-0047

Speaker

Edward Shields (Weizmann Institute of Science (IL))

Description

Machine Learning methods are becoming more prevalent in physics for a variety of tasks, and some have already been demonstrated within the directional recoil detection community for reconstruction and direction prediction tasks. We will propose a modern architecture to predict direction and class of recoils in one model. Further, we will propose a generative ML model, based on a super-resolution technique, to enhance detector readout beyond the physical readout of a TPC. If proven to be viable, this technique could be incorporated into detector designs, to allow for greater readout granularity beyond financial and technical constraints.

Presentation materials