neural network data operation method based on a NOR Flash module

Abstract

The invention discloses a neural network data operation method based on a NOR Flash module, and the method comprises an input signal generation circuit, an output signal measurement circuit, and a storage array composed of a plurality of storage units or current mirrors, and is characterized by comprising the following steps of equipping the NOR Flash module with at least one storage array; setting a threshold voltage of each storage unit in the storage array according to a preset weight value in the neural network; converting the to-be-operated data information into a set format; inputting the converted to-be-operated data information to the input end of the storage array through an input signal generation circuit; enabling the storage array to perform operation processing through a preset algorithm according to the threshold voltage of the storage unit and data parameters input to the input end of the storage array through an input signal generation circuit; and outputting an outputresult obtained after operation processing of the storage array through the output signal measurement circuit.

Description

technical field [0001] The invention relates to the field of semiconductor integrated circuits, in particular to a data operation method of a neural network based on a NOR Flash module. Background technique [0002] Flash memory is widely used as a non-volatile storage technology. The usual implementation method is to use a floating gate (Floating Gate) or a charge trap structure (Charge Trap) to store charges in a field effect transistor (Field Effect Transistor, FET) to form a storage unit. , storage cells form an array for storing large amounts of data. [0003] According to different array structures, flash memory can be divided into two types: NAND type and NOR type. The storage cells of NOR flash memory (NOR Flash) are connected in parallel. NOR Flash can perform independent read and write operations for each of its storage units, and has lower read latency than NAND Flash. [0004] As an artificial intelligence (AI) algorithm, neural network is widely used in many ...

AI Technical Summary

Problems solved by technology

[0005] Large-scale neural networks, with their huge parameter sets and calculations, and their complex network types (CNN, RNN, DNN, etc.), have brought many obstacles to the design of AI chips

Two of the problems are particularly obvious. One is the limitation of computing bandwidth on computing speed, and the other is the requirement of frequent data access to power consumption.

Images