Looking for breakthrough ideas for innovation challenges? Try Patsnap Eureka!

Serialization task completion method and system based on memory consolidation mechanism and GAN model

A task completion and serialization technology, applied in neural learning methods, biological neural network models, computer components, etc., can solve problems such as inability to complete serialization tasks, model forgetting, etc.

Pending Publication Date: 2021-01-22
CAS HEFEI INST OF TECH INNOVATION +2
View PDF0 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0007] The purpose of the present invention is to solve the defect that the model will forget in the multi-task scene in the prior art, so that the serialization task cannot be completed, and provide a method and system for completing the serialization task based on the memory consolidation mechanism and the GAN model to solve the above problems

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
  • Serialization task completion method and system based on memory consolidation mechanism and GAN model
  • Serialization task completion method and system based on memory consolidation mechanism and GAN model
  • Serialization task completion method and system based on memory consolidation mechanism and GAN model

Examples

Experimental program
Comparison scheme
Effect test

Embodiment approach

[0125] Here, as the first embodiment of the present invention, when it is the task of continuous image generation, the neural network often faces the situation of insufficient data volume during the training process. The insufficient data volume will cause the network to be unable to fully learn the characteristics of the data, resulting in Model overfitting problem. One way to solve the insufficient amount of data is to expand the data set, and using GAN to generate the data set is an effective and simple method. We apply the GAN continuous learning system to image generation and test whether it has the ability to continuously generate multi-category images. In the experiment, the MNIST dataset is used, which contains handwritten digits 0, 1, 2, 3, 4, 5, 6, 7, 8 and 9. We divide it into two groups, the first group is 0-4 and the second group is 5-9. Task 1 is to generate the first set of numbers and task 2 is to generate the second set of numbers. Specific steps are as fol...

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 relates to a serialization task completion method based on a memory consolidation mechanism and a GAN model. Compared with the prior art, the defect that a serialization task cannot be completed due to the fact that a model is forgotten in a multi-task scene is overcome. The method comprises the following steps: obtaining a serialized task; setting an indexer and generating a task index; performing task training by using a GAN model; performing joint training of pseudo samples; and completing the new serialization task. Through protection of important parameters and design of memory playback, a memory consolidation mechanism is applied to the GAN model, so that the GAN model has a multi-task processing capability, important information in sub-tasks can be reserved, non-important information is forgotten, and completion of serialized tasks is realized.

Description

technical field [0001] The present invention relates to the technical field of serialized task processing methods, in particular to a serialized task completion method and system based on a memory consolidation mechanism and a GAN model. Background technique [0002] Generative Adversarial Networks (GAN, Generative Adversarial Networks) is a deep learning model, which consists of a generator and a discriminator, trained by confrontational learning, the purpose is to estimate the potential distribution of data samples and generate new data sample. The purpose of the generator is to learn the real data distribution as much as possible, and the purpose of the discriminator is to correctly judge whether the input data comes from the real data or the generator; in order to win the game, these two game participants need to continuously optimize and improve each One's own generation ability and discrimination ability, this learning optimization process is to find a Nash equilibriu...

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
IPC IPC(8): G06K9/62G06N3/08G10L21/003G06F17/15B33Y40/00B29C64/393
CPCG06N3/08B29C64/393B33Y40/00G10L21/003G06F17/15B29L2031/443B29L2031/7158G10L2021/0135G06F18/214
Inventor 常一帆刘鹏程王亦凡孙晓晴刘文涛李文波王矿王伟祥邱骐
Owner CAS HEFEI INST OF TECH INNOVATION
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Patsnap Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Patsnap Eureka Blog
Learn More
PatSnap group products

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

© 2024 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap|About US| Contact US: help@patsnap.com