High-dimensional sparse image characterization method based on deep semi-supervised learning framework

The invention discloses a high-dimensional sparse image representation method based on a deep semi-supervised learning framework. The method comprises the following steps: 1) initializing a Semi-DRRBM model parameter theta; 2) obtaining features of a sample hidden layer; 3) using reconstructed visib...

Full description

Saved in:
Bibliographic Details
Main Authors LIU SURUI, XIANG DAO, HU JIANGTAO, ZHOU TUNAN, XIAO YU, CHU JIELEI, SHAN JUNPENG, YOU ZEQING, YU HUA
Format Patent
LanguageChinese
English
Published 18.07.2023
Subjects
CALCULATING
COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS
COMPUTING
COUNTING
ELECTRIC DIGITAL DATA PROCESSING
PHYSICS
Online AccessGet full text

Cover

More Information
Summary:The invention discloses a high-dimensional sparse image representation method based on a deep semi-supervised learning framework. The method comprises the following steps: 1) initializing a Semi-DRRBM model parameter theta; 2) obtaining features of a sample hidden layer; 3) using reconstructed visible layer data; 4) obtaining hidden layer features of sample reconstruction; 5) calculating H (theta; g) and H '(theta; g '); (6) obtaining the calculation of the current time by using the previous hidden layer data h and visual layer data v, the reconstructed hidden layer data h'and visual layer data v 'and the parameters obtained in the ith calculation; (7) after a new parameter theta is obtained in the step (6), taking the new parameter theta as a new input, and carrying out iteration in the steps (2)-(6) again; and 8) returning a parameter theta of the Semi-DRRBM model, and substituting the parameter theta into the Semi-DRRBMRBM model to complete dimension reduction. According to the method, low-dimensional and
Bibliography:Application Number: CN202310422895