Numerical Stability in Deep Learning

This RICAM workshop examines the numerical foundations of reliable deep-learning computation. Low-precision and mixed-precision arithmetic, quantization and parallel GPU computation enable larger models and deployment on constrained devices, while introducing questions about error, stability and performance. Researchers from theoretical and applied communities will compare methods for stable training and efficient inference and identify directions for further work. The programme brings together expertise in numerical analysis, machine learning and hardware-aware computation.

The in-person workshop runs on 2–6 November at RICAM, Science Park 2, fourth floor, Altenberger Straße 69, Linz. Monday registration and welcome run 14:00–14:40, followed by talks until 17:00. Tuesday through Friday sessions begin at 09:00; Friday’s programme ends at 11:40. Times are Europe/Vienna, CET (UTC+01:00). Register through the special-semester form, select Workshop 2 and Regular Participant, and use an institutional email address to verify affiliation. Registration is binding; the organizer asks for cancellation at least one week before the workshop and states that unreported non-attendance can exclude participants from future events.