A method and system for accelerated computing of integrated circuits based on convolutional neural network algorithm

Abstract

The invention belongs to the field of artificial intelligence technology, and in particular relates to a method and system for accelerated calculation of an integrated circuit based on a convolutional neural network algorithm, wherein the system includes inputting convolution kernel data and external data in parallel from different directions to the multiplication and accumulation Each multiplication accumulator unit in the multiplication accumulator unit formation simultaneously performs corresponding multiplication and accumulation processing on the convolution kernel data and external data flowing through it in parallel, and outputs them to the data storage unit respectively. The present invention solves the problem of Due to the existence of the existing technology, a disadvantage of the convolutional neural network is that it has a huge amount of calculation, and it is not easy to perform real-time calculations on integrated circuits or embedded devices. Traditional processors based on serial architectures are not easy to meet the requirements. Therefore, how to Quickly completing convolutional neural network operations is an important problem to be solved. The present invention has the characteristics of less reading times, high computing throughput, and low bandwidth requirements, and has the beneficial technical effect of greatly improving the real-time performance of convolutional neural network operations.

Description

technical field [0001] The invention belongs to the technical field of artificial intelligence, and in particular relates to a method for accelerated calculation of an integrated circuit based on a convolutional neural network algorithm. At the same time, the invention also provides a system for accelerated calculation of an integrated circuit based on a convolutional neural network algorithm. Background technique [0002] Convolutional neural network is a kind of feedforward neural network, which is often used in image recognition. It generally includes convolutional layer, pooling layer and fully connected layer. The convolution operation of convolutional layer is that each weight in the convolution kernel Multiply the corresponding input data point-to-point, and then accumulate the point multiplication results to obtain an output data. After that, according to the step size setting of the convolution layer, slide the convolution kernel and repeat the above operation. One ...

AI Technical Summary

Problems solved by technology

[0003] The present invention provides a method and system for accelerated calculation of an integrated circuit based on a convolutional neural network algorithm to solve the problems in the prior art proposed in the above-mentioned background technology. Due to a disadvantage of the convolutional neural network, its computational complexity is huge. It is not easy to perform real-time calculations on embedded devices, and traditional processors based on serial architectures are not easy to meet the requirements. Therefore, how to quickly complete convolutional neural network operations is an important problem that needs to be solved.

Images