New ML method speeds up uncertainty quantification for particle accelerators

Researchers have developed a novel empirical Bayes method to efficiently tune hyperparameters for flexible, heteroscedastic Spectral-normalized Neural Gaussian Processes. This technique significantly reduces computational costs associated with complex machine learning models, making them more feasible for large-scale scientific applications. The method was successfully applied to predict the dynamic aperture in particle accelerators, specifically for the Large Hadron Collider at CERN, offering a more cost-effective alternative to traditional simulation-heavy approaches while maintaining competitive performance and accurate uncertainty estimates. AI

IMPACT This method could enable more efficient uncertainty quantification in complex scientific simulations, reducing computational costs for research in fields like particle physics.

RANK_REASON Academic paper detailing a new machine learning method. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.LG →

• arXiv
• CERN
• Ekaterina Krymova
• Large Hadron Collider
• Spectral-normalized Neural Gaussian Process

• paper
• infra

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →