Neural Network Systems and Neural Network Computation Methods

CN113139641BActive Publication Date: 2026-06-30HUAWEI TECH CO LTD +1
View PDF 2 Cites 0 Cited by

Patent Information

Application Number
CN202010067674.9
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2020-01-20
Publication Date
2026-06-30
Estimated Expiration
2040-01-20

AI Technical Summary

Technical Problem

In artificial neural network chips, DACs and ADCs occupy a large proportion of the area and power consumption, affecting chip performance improvement.

Method used

Using resistive random access memory (RRAM) as an analog device, multiplication and addition operations are achieved through memristor arrays and current-to-voltage converters, avoiding the use of DACs and ADCs. Neuron functions are realized by utilizing the resistance changes of RRAM, and differential operations are performed through differential circuits.

Benefits of technology

It saves chip area and power consumption, while improving the feature selectivity and recognition ability of the neural network system and enhancing computational efficiency.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN113139641B_ABST
    Figure CN113139641B_ABST
Patent Text Reader

Abstract

This application provides a neural network system and a neural network calculation method, relating to the field of electronic technology, which can save chip area and power consumption. The neural network system includes a first memristor array, multiple current-to-voltage converters, and multiple resistive random access memories (RRAMs). The first memristor array includes multiple rows and columns of memristors for receiving first input data and performing multiplication and addition operations on the first input data according to configured first weights to obtain current. Each current-to-voltage converter is connected to the output terminal of at least one column of memristors in the first memristor array, for receiving the current of the corresponding column of memristors and converting the received current into voltage. Each RRAM is connected to at least one current-to-voltage converter, for receiving the voltage of the corresponding current-to-voltage converter, and obtaining one of the calculation results from the first calculation results based on the received voltage and the initial resistance state of the corresponding RRAM. The initial resistance state of each RRAM is a high resistance state.
Need to check novelty before this filing date? Find Prior Art

Citation Information

Patent Citations

  • Memory device and data-processing method based on multi-layer RRAM crossbar array

    CN108475522A

  • A convolution neural network-on-chip learning system based on nonvolatile memory

    CN109460817A