Score: 1

Convergence for Discrete Parameter Updates

Published: December 3, 2025 | arXiv ID: 2512.04051v1

By: Paul Wilson, Fabio Zanasi, George Constantinides

Potential Business Impact:

Makes computer learning faster and use less power.

Business Areas:
Machine Learning Artificial Intelligence, Data and Analytics, Software

Modern deep learning models require immense computational resources, motivating research into low-precision training. Quantised training addresses this by representing training components in low-bit integers, but typically relies on discretising real-valued updates. We introduce an alternative approach where the update rule itself is discrete, avoiding the quantisation of continuous updates by design. We establish convergence guarantees for a general class of such discrete schemes, and present a multinomial update rule as a concrete example, supported by empirical evaluation. This perspective opens new avenues for efficient training, particularly for models with inherently discrete structure.

Country of Origin
🇬🇧 United Kingdom

Page Count
10 pages

Category
Computer Science:
Machine Learning (CS)