USPTO Patent Grant: Integrated Memory System for Neural Networks

ChangeBridge: Patent Grants - AI & Computing (G06N)

Published March 24th, 2026

Detected March 25th, 2026

Summary

The USPTO has granted patent US12585927B2 for an integrated memory system for neural networks, developed by The Board of Trustees of the University of Illinois. The patent covers a memory module designed to compute, store, and sample neural network weights, optimizing precision for machine learning data streams.

View original document View source feed page

What changed

The United States Patent and Trademark Office (USPTO) has issued patent US12585927B2, titled 'Integrated memory system for high performance Bayesian and classical inference of neural networks.' The patent, assigned to The Board of Trustees of the University of Illinois, details a memory module system designed for neural networks processing machine learning data streams. Key features include embedded random number generators and adaptive operating precision to optimize computing effort.

This patent grant represents a new intellectual property asset in the field of AI and computing hardware. While it does not impose new regulatory obligations on businesses, it signifies innovation in memory systems for AI applications. Companies operating in the AI and semiconductor sectors may wish to review the patent's claims to understand the scope of the protected technology and assess potential impacts on their own research and development or product strategies.

Source document (simplified)

← USPTO Patent Grants

Integrated memory system for high performance Bayesian and classical inference of neural networks

Grant US12585927B2 Kind: B2 Mar 24, 2026

Assignee

THE BOARD OF TRUSTEES OF THE UNIVERSITY OF ILLINOIS

Inventors

Amit Ranjan Trivedi, Theja Tulabandhula, Priyesh Shukla, Ahish Shylendra, Shamma Nasrin

Abstract

A memory module system for a high-dimensional weight space neural network configured to process machine learning data streams using Bayesian Inference and/or Classical Inference is set forth. The memory module can include embedded high speed random number generators (RNGs). The memory module is configured to compute, store and sample neural network weights by adapting operating precision to optimize the computing effort based on available weight space and application specifications.

CPC Classifications

G06G 7/16 G06N 3/04 G06N 3/045 G06N 3/047 G06N 3/063 G06N 3/065 G06N 3/08 G06F 15/7821 G06F 15/785 G06F 7/58 G06F 7/582 G06F 7/588 G06F 17/15 G06F 17/153 G06F 17/16 G11C 11/419 G11C 11/54 G11C 7/1006