Mitsui has filed a patent for a neural network generating device that can generate a neural network execution model based on hardware and network information. The device includes an execution model generation unit and a learning unit to generate the trained parameters of the model. GlobalData’s report on Mitsui gives a 360-degree view of the company including its patenting strategy. Buy the report here.

Smarter leaders trust GlobalData

Report-cover

Data Insights Mitsui & Co Ltd - Company Profile

Buy the Report

Data Insights

The gold standard of business intelligence.

Find out more

According to GlobalData’s company profile on Mitsui, Hydrogen storage alloys was a key innovation area identified from patents. Mitsui's grant share as of September 2023 was 39%. Grant share is based on the ratio of number of grants to total number of patents.

Neural network generating device for generating neural network execution model

Source: United States Patent and Trademark Office (USPTO). Credit: Mitsui & Co Ltd

A recently filed patent (Publication Number: US20230316071A1) describes a neural network generating device that is capable of generating a neural network execution model for operating a neural network. The device includes an execution model generation unit that generates the neural network execution model based on hardware information about the hardware in which the model will operate, as well as network information about the neural network itself. Additionally, the device includes a learning unit that generates learned parameters for the generated neural network execution model.

The neural network generating device may also include a hardware generation unit that generates a neural network hardware model based on the hardware information and the neural network execution model. This allows for a comprehensive understanding of the hardware requirements for operating the neural network.

The execution model generation unit is capable of partitioning convolution operations within the neural network execution model based on the generated model. This allows for efficient processing and optimization of the neural network.

The learning unit performs associated operations with higher precision than those implemented by the neural network execution model when generating the learned parameters. This ensures that the learned parameters are accurate and reliable.

The neural network execution model itself consists of a convolution operation circuit for implementing convolution operations and a quantization operation circuit for implementing quantization operations. The learning unit is capable of performing convolution operations with higher precision than those implemented by the convolution operation circuit. Additionally, the learning unit learns quantization parameters used by the quantization operation circuit for the quantization operations. This allows for improved accuracy and efficiency in the neural network's operations.

The patent also describes a neural network generating method that follows a similar process to the neural network generating device. The method includes steps for acquiring hardware information, setting network information, generating the neural network execution model, and learning the learned parameters.

Furthermore, the patent includes a non-transitory computer-readable recording medium that stores a neural network generating program. This program allows a computer to generate a neural network execution model based on hardware information and network information, as well as learn the learned parameters.

Overall, this patent describes a neural network generating device, method, and program that aim to optimize the performance and accuracy of neural networks by considering hardware information, partitioning operations, and performing associated operations with higher precision.

To know more about GlobalData’s detailed insights on Mitsui, buy the report here.

Data Insights

From

The gold standard of business intelligence.

Blending expert knowledge with cutting-edge technology, GlobalData’s unrivalled proprietary data will enable you to decode what’s happening in your market. You can make better informed decisions and gain a future-proof advantage over your competitors.

GlobalData

GlobalData, the leading provider of industry intelligence, provided the underlying data, research, and analysis used to produce this article.

GlobalData’s Company Filings Analytics uses machine learning to uncover key insights and track sentiment across millions of regulatory filings and other corporate disclosures for thousands of companies representing the world’s largest industries. This analysis is combined with crucial details on strategic and investment priorities, innovation strategies, and CXO insights to provide comprehensive company profiles.