Samsung Patent for Neural Network Convolution Operation

ChangeBridge: Patent Grants - AI & Computing (G06N)

Published March 24th, 2026

Detected March 25th, 2026

Summary

The USPTO has granted Samsung Electronics Co., Ltd. a patent (US12585913B2) for a method and apparatus related to neural network convolution operations. The patent details a processor-implemented method for generating output features by performing convolution operations and efficiently storing them in memory.

View original document View source feed page

What changed

The United States Patent and Trademark Office (USPTO) has granted patent US12585913B2 to Samsung Electronics Co., Ltd. The patent covers a processor-implemented method for performing neural network convolution operations. Specifically, it describes generating an output feature map by convolving input feature map data with weight kernels and efficiently storing the results by overwriting memory space.

This patent grant is primarily an intellectual property development and does not impose new regulatory obligations on businesses. However, it signifies an advancement in AI hardware and software optimization, particularly for companies involved in developing or utilizing neural networks. Compliance officers in the technology sector should be aware of such patent grants as they can influence competitive landscapes and future technological standards.

Source document (simplified)

← USPTO Patent Grants

Method and apparatus with neural network convolution operation

Grant US12585913B2 Kind: B2 Mar 24, 2026

Assignee

Samsung Electronics Co., Ltd.

Inventors

Jinwoo Son, Changyong Son, Jaehyoung Yoo, Seohyung Lee, Sangil Jung, Changin Choi

Abstract

A processor-implemented neural network method includes: generating a first output line of an output feature map by performing a convolution operation between a first input line group of an input feature map and weight kernels; generating a first output of an operation block including the convolution operation based on the first output line; and storing the first output in a memory in which the input feature map is stored by overwriting the first output to a memory space of the memory.

CPC Classifications

G06N 3/04 G06N 3/08 G06N 3/0464 G06N 3/0475 G06N 3/0495 G06N 3/0499 G06N 3/0895 G06N 3/09 G06N 3/091 G06N 3/092 G06N 3/094 G06N 3/096 G06N 3/098 G06N 3/0985