Handwritten numeral generation method based on double-discriminator weighted hybrid generative adversarial network

A dual discriminator and discriminator technology, applied in biological neural network models, neural learning methods, instruments, etc., can solve problems such as inability to use handwritten digit generation methods, mode collapse, gradient disappearance, etc., to solve the gradient disappearance. Issues, effects to fix modal crash issues

Pending Publication Date: 2022-07-15
XIAN UNIV OF SCI & TECH +1
View PDF0 Cites 1 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

MGAN only uses JS divergence, which is prone to gradient disappearance and causes model collapse
[0006] None of the above methods work well for handwritten digit generation methods

Method used

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View more

Image

Smart Image Click on the blue labels to locate them in the text.
Viewing Examples
Smart Image
  • Handwritten numeral generation method based on double-discriminator weighted hybrid generative adversarial network
  • Handwritten numeral generation method based on double-discriminator weighted hybrid generative adversarial network
  • Handwritten numeral generation method based on double-discriminator weighted hybrid generative adversarial network

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0126] like figure 1 As shown, the handwritten digit generation method based on the dual discriminator weighted hybrid generative adversarial network of the present invention, the method comprises the following steps:

[0127] Step 1: Obtain the real handwritten digital image of the target user;

[0128] Step 2: Input the real handwritten digital image of the target user into the pre-trained dual discriminator weighted hybrid generative adversarial network (D2WMGAN network); wherein, the training process of the dual discriminator weighted hybrid generative adversarial network (D2WMGAN network) is as follows: :

[0129] Step 201, using the MNIST data set as a training sample;

[0130] The MNIST dataset consists of handwritten digits from 250 different people, 50% of whom are high school students and 50% from the Census Bureau staff. The MNIST dataset contains 70,000 pictures of handwritten digits, of which 60,000 are The training set, 10,000 as the test set, is divided into ...

Embodiment 2

[0179] The difference between this embodiment and Embodiment 1 is that the training process of the dual discriminator weighted hybrid generative adversarial network (D2WMGAN network) described in step 2 further includes step 205, which is to perform the training of the double discriminator weighted hybrid generative adversarial network (D2WMGAN network). Network) to carry out theoretical analysis and verification, and verify that under the optimal discriminator, the generator generates real handwritten digital data by minimizing the KL divergence and the reverse KL divergence between the generated data and the real data.

[0180] In this embodiment, in step 205, the trained dual discriminator weighted hybrid generative adversarial network (D2WMGAN network) is theoretically analyzed and verified to verify that under the optimal discriminator, by minimizing the difference between the generated data and the real data KL divergence and reverse KL divergence, the specific process fo...

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to view more

PUM

No PUM Login to view more

Abstract

The invention discloses a handwritten numeral generation method based on a double-discriminator weighted hybrid generative adversarial network. The method comprises the following steps: 1, acquiring a real handwritten numeral image of a target user; 2, inputting the real handwritten digital image of the target user into a pre-trained double-discriminator weighted hybrid generative adversarial network; and step 3, the double-discriminator weighted hybrid generative adversarial network processes the real handwritten digital image of the target user to obtain handwritten digital data. According to the method, the advantages of multiple generators and double discriminators are combined, an objective function is reconstructed, a model structure is designed, and the phenomenon of gradient disappearance is avoided from the angles of a network model structure and a loss function; the advantages of the forward KL divergence and the reverse KL divergence are combined, so that the generated modes are diversified, and the problem of mode collapse of the GAN is improved.

Description

technical field [0001] The invention belongs to the technical field of handwritten digit generation methods, and in particular relates to a handwritten digit generation method based on a dual discriminator weighted hybrid generative confrontation network. Background technique [0002] Generative Adversarial Networks (GAN) is an adversarial learning method developed in recent years. GAN is composed of a generator G and a discriminator D, using the idea of ​​game theory, the two play against each other, and the goal is to find the Nash equilibrium (Nash equilibrium) in the continuous non-convex problem with high-dimensional parameters. GAN has been proven to generate realistic images, which is very helpful in data enhancement and image completion, and is mainly used in image super-resolution reconstruction, transfer learning, image inpainting and other fields. [0003] But given an optimal discriminator, the loss function of the generator is equivalent to minimizing the real ...

Claims

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to view more

Application Information

Patent Timeline
no application Login to view more
Patent Type & Authority Applications(China)
IPC IPC(8): G06N3/04G06N3/08G06K9/62
CPCG06N3/084G06N3/047G06N3/045G06F18/214
Inventor 刘宝王良宋美玉翟晓航张金玉
Owner XIAN UNIV OF SCI & TECH
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Try Eureka
PatSnap group products

Browse by: Latest US Patents, China's latest patents, Technical Efficacy Thesaurus, Application Domain, Technology Topic.

© 2024 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap