Data processing method, data processing device and chip

Abstract

The embodiment of the invention discloses a data processing method, a data processing device and a chip. Wherein, the data processing method includes: in the process of reverse processing the N-layer neural network, detecting whether the input data of the i-th layer network to be processed is currently stored; wherein, N is an integer greater than 1, and i is an integer, And 1<i≤N; if the input data of the i-th layer network is not stored, the input data of the i-1 layer network is processed through the i-1 layer network to obtain the input data of the i-th layer network; after that, By default, based on the output related data of the i-th layer network and the input data of the i-th layer network, obtain the training data of the i-1 layer network; otherwise, directly execute the preset method based on the i-th layer The output related data of the network and the input data of the i-th layer network are used to obtain the training data of the i‑1th layer network. The embodiment of the present invention can save the bandwidth of data storage.

Description

technical field [0001] The invention belongs to the field of data processing, and in particular relates to a data processing method, a data processing device and a chip. Background technique [0002] In recent years, with the rapid development of deep learning technology, the representative technology in deep learning technology - Convolutional Neural Networks (CNN) has been widely used in the field of pattern recognition such as image recognition and speech recognition. [0003] Convolutional neural network is a multi-layer neural network, each layer is composed of multiple two-dimensional planes, and each plane is composed of multiple independent neurons. It is essentially a mapping from input to output, and can learn a large number of mapping relationships between input and output. As long as the convolutional neural network is trained with a known pattern, the convolutional neural network has input and output pairs. mapping capability between them. [0004] At present,...

AI Technical Summary

Problems solved by technology

With the increasing size of the convolutional neural network, the number of layers of the network and the number of neurons in each layer continue to increase, it is difficult to store all the data of the entire convolutional neural network on the same chip, and if these data are stored on the chip In external storage devices, due to the huge amount of interactive data, the bandwidth between the chip and the storage device will limit the data transmission efficiency, thus affecting the training efficiency of the convolutional neural network.

Images