An interpretable unsupervised representation learning for high precision measurement in particle physics
Researchers have developed a new unsupervised representation learning network called Histogram AutoEncoder (HistoAE) for high-precision measurements in particle physics. This model features a custom histogram-based loss function that structures the latent space, making it physically interpretable. When applied to silicon microstrip detectors, HistoAE achieved charge resolution of 0.25e and position resolution of 3μm on beam-test data, matching conventional methods. The network's generative capabilities also allow for fast detector simulations. AI
IMPACT This new AI model offers a path to more precise and interpretable measurements in particle physics, potentially accelerating research and enabling faster simulations.