Title:
INTERPRETABLE MULTILAYER INFORMATION SCREENING NETWORK
Document Type and Number:
WIPO Patent Application WO/2021/232172
Kind Code:
A1
Abstract:
In the application of the present invention, a working principle that a multilayer neural network is used for information screening is provided, and a visual intermediate layer information processing method is provided according to the principle, so that the reliability of information processing of the multilayer neural network is improved. In the application of the present invention, a method for performing multiresolution pre-processing on data, and then screening information by using the multilayer neural network is also provided. Also provided is a method for performing multiresolution extraction on a dynamic characteristic. Moreover, these methods lay a foundation for further improving the generalization capability of models.
Inventors:
CHEN YONGCONG (CN)
Application Number:
PCT/CN2020/000110
Publication Date:
November 25, 2021
Filing Date:
May 18, 2020
Export Citation:
Assignee:
CHEN YONGCONG (CN)
International Classes:
G06N3/04; G06K9/62; G06N3/08
Foreign References:
| CN109902623A | 2019-06-18 | |||
| CN110059796A | 2019-07-26 | |||
| CN105095862A | 2015-11-25 | |||
| CN107862275A | 2018-03-30 | |||
| CN110110719A | 2019-08-09 | |||
| CN111062297A | 2020-04-24 | |||
| US20190385021A1 | 2019-12-19 |
Download PDF: