Title:
CONVOLUTIONAL NEURAL NETWORK-BASED DATA PROCESSING METHOD AND DEVICE
Document Type and Number:
WIPO Patent Application WO/2020/258568
Kind Code:
A1
Abstract:
A convolutional neural network-based data processing method. For any convolutional layer in a convolutional neural network, calculation is performed, by means of a convolution kernel of the convolutional layer, on elements of data input to the convolutional layer one by one so as to obtain convoluted values of the respective elements. Each calculation obtains a convoluted value, and this convoluted value and convoluted values of elements in the same region obtained by calculation by the same convolution kernel are added together to obtain an output element of the convolution kernel corresponding to the region. According to the data processing method, in a process of calculating a convoluted value, each calculation obtains a convoluted value, and this convoluted value is added to a corresponding convoluted sum to directly obtain an element in an output of a convolutional layer. In this way, an output of a convolutional layer can be obtained after all convoluted values have been calculated without having to read convoluted values from a storage apparatus. Since interaction with a storage apparatus is effectively reduced in a process of calculating an output of a convolutional layer, data processing efficiency can be enhanced.
Inventors:
MEI GUOQIANG (CN)
HAO RUI (CN)
HAO RUI (CN)
Application Number:
PCT/CN2019/108928
Publication Date:
December 30, 2020
Filing Date:
September 29, 2019
Export Citation:
Assignee:
INSPUR SUZHOU INTELLIGENT TECHNOLOGY CO LTD (CN)
International Classes:
G06F17/16
Foreign References:
| CN109086244A | 2018-12-25 | |||
| CN106951395A | 2017-07-14 | |||
| CN108205702A | 2018-06-26 | |||
| US20170208081A1 | 2017-07-20 |
Attorney, Agent or Firm:
UNITALEN ATTORNEYS AT LAW (CN)
Download PDF: