Self-Organised Polylinear Regression Maps As a Method of Resolving HPGe Detector Responses

Within the field of Nuclear Instrumentation, the technique of Pulse Shape Analysis (PSA) is used to apply bespoke timing and energy corrections for detectors by interrogating the signal profiles corresponding to their spatial response. Parametric PSA methods are suitable for high-rate applications but rely on human-interpretable corrections and are often limited in their accuracy and application to multi-interaction signals. Non-parametric PSA methods such as the gamma-ray localisation used in the Advanced Gamma Tracking Array (AGATA) provide excellent prediction accuracy but are rate-limited and can only operate at a few kHz. In this work we propose a pseudo-parametric method using Self-Organising Maps (SOMs) for generating a human-interpretable approximation of the underlying detector response. These SOMs can be experimentally trained for accurate analytical regression of a variety of parameters from radial and cartesian position to timing. This method provides an efficient and tuneable way for the prediction of signal characteristics when a detector exhibits complex non-linear behaviour whilst capable of being implemented real-time on conventional digitiser hardware.