Acceleration processing unit based on convolutional neural network and array structure thereof

Abstract

The invention discloses an acceleration processing unit based on a convolutional neural network, and is used for performing convolution operation on local data. The local data include multiple multimedia data. The acceleration processing unit comprises a first register, a second register, a third register, a fourth register, a fifth register, a multiplier, an adder, a first multipath selector and a second multipath selector. The single acceleration processing unit controls the first multipath selector and the second multipath selector so that the multiplier and the adder are enabled to be repeatedly usable, one acceleration processing unit is enabled to only need one multiplier and one adder to complete convolution operation and use of the multiplier and the adder can be reduced. Processing speed can be enhanced and energy consumption can be reduced by reducing use of the multiplier and the adder in implementing the same convolution operation, and the on-chip area of the single acceleration processing unit is smaller.

Description

technical field [0001] The invention relates to a convolutional neural network, in particular to an accelerated processing unit and an array structure in a convolutional layer of the convolutional neural network. Background technique [0002] Compared with shallow learning, deep learning (dee acceleration processing unit Plea2ning) means that the machine learns the rules from historical data through algorithms, and makes intelligent identification and prediction of things. [0003] Convolutional Neural Network (Convolutional Neu2al Netwo2k, CNN) is a kind of dee accelerated processing unit Plea2ning netwo2k, which was invented in the early 1980s. It is composed of artificial neurons arranged in multiple layers. The convolutional neural network reflects the human brain processing vision. Methods. As Moore's Law drives computer technology to become more and more powerful, convolutional neural networks can better imitate the actual operation of biological neural networks, avoi...

AI Technical Summary

Problems solved by technology

The existing accelerated processing unit is designed with an adder and a multiplier for each input multimedia data. When the accelerated processing unit needs to process multiple local data, it means that each accelerated processing unit includes multiple additions. multiplier and multiple multipliers, this design leads to a large area of ​​the accelerated processing unit chip, high power consumption, and the processing speed needs to be improved

Images