Huawei Patent for Parameter Optimization Method Granted

ChangeBridge: Patent Grants - AI & Computing (G06N)

Published March 24th, 2026

Detected March 25th, 2026

Summary

The USPTO has granted Huawei Technologies Co., Ltd. a patent for a parameter optimization method applicable to neural network models. The patent, filed in 2022, describes a method for adjusting model parameters based on segmented Kronecker factor matrices.

View original document View source feed page

What changed

The United States Patent and Trademark Office (USPTO) has granted patent US12585945B2 to Huawei Technologies Co., Ltd. for a parameter optimization method. This patent, filed on November 29, 2022, details a process for adjusting parameters in neural network models by obtaining and segmenting a Kronecker factor matrix to derive submatrices, which are then used for parameter adjustment.

This grant represents a new intellectual property asset for Huawei in the field of AI and computing. While patents do not impose direct compliance obligations on other entities, they can influence competitive landscapes and licensing strategies. Companies developing or utilizing similar parameter optimization techniques in neural networks should be aware of this granted patent and consider its implications for their own IP strategy and potential licensing needs.

Source document (simplified)

← USPTO Patent Grants

Parameter optimization method, electronic device, and storage medium

Grant US12585945B2 Kind: B2 Mar 24, 2026

Assignee

Huawei Technologies Co., Ltd.

Inventors

Ningxi Ni, Min Wang, Zidong Wang, Fan Yu, Lei Chen

Abstract

Example parameter optimization methods and apparatus are described. In one example parameter optimization method, a data processing device obtains a Kronecker factor matrix that is used to indicate a higher-order information matrix of a neural network model, and segments the Kronecker factor matrix to obtain a plurality of square matrices. The obtained plurality of square matrices are submatrices of the Kronecker factor matrix, and main diagonal lines of the plurality of square matrices each are in a one-to-one correspondence with a part of a main diagonal line of the Kronecker factor matrix. Then, the data processing device adjusts a parameter of the neural network model based on the obtained plurality of square matrices.