Eureka AIR delivers breakthrough ideas for toughest innovation challenges, trusted by R&D personnel around the world.

Decomposition and synthesis method and system for deep learning neural networks

A neural network and deep learning technology, applied in neural learning methods, biological neural network models, neural architectures, etc., can solve the problems of complex training, poor training and prediction effects, etc.

Active Publication Date: 2018-07-20
SUPERPOWER INNOVATION INTELLIGENT TECH DONGGUAN CO LTD
View PDF3 Cites 4 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

If the input data variable is data at multiple time points, the training will be very complicated, and will lead to poor training and prediction effects

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
  • Decomposition and synthesis method and system for deep learning neural networks
  • Decomposition and synthesis method and system for deep learning neural networks
  • Decomposition and synthesis method and system for deep learning neural networks

Examples

Experimental program
Comparison scheme
Effect test

preparation example Construction

[0047] In conjunction with the accompanying drawings, a method for decomposing and synthesizing a deep learning neural network of the present invention comprises the following steps:

[0048] Step 1. Obtain input data variable set A and output data variable set B;

[0049] Step 2. According to the correlation between the input data variables in the input data variable set A, cluster the input data variables, cluster the input data variables in the input data variable set A into different subsets, and obtain N subsets ; The N is greater than or equal to 1;

[0050] Step 3, initialize a corresponding first deep learning neural network for each subset in the N subsets, obtain N first deep learning neural networks, and then use all input data variables in each subset as the corresponding The input data variable of the first deep learning neural network;

[0051] Step 4, initialize a second deep learning neural network, the input layer nodes of the second deep learning neural net...

Embodiment

[0086] to combine figure 1 , the decomposition and synthesis method of deep neural network of the present invention, comprise the following steps:

[0087] Step 1. Obtain the input data variable set A as the pixel matrix of the full-body photo, and the output data variable set B as gender, age, and height.

[0088] Step 2. According to the correlation between the input data variables in the input data variable set A, use the k-means method to cluster the input data variables, calculate the distance between different pixels during clustering, and try to make each pixel in the same subset The distance between pixels is short, but the distance between pixels between different subsets is long. For example, after clustering the input data variable set A, three subsets are obtained, which are head pixel matrix H, upper body pixel matrix U, and lower body pixels matrixD. Among them, the upper body refers to the body parts above the waist, and the lower body refers to the body parts...

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 decomposition and synthesis method and system for deep learning neural networks. The method comprises a step of firstly acquiring an input data variable set and an output data variable set, a step of decomposing the input data variable set into N subsets, a step of initializing first deep learning neural networks, a step of initializing a second deep learning neural network, a step of initializing a third deep learning neural network, a step of carrying out unsupervised training on the first deep learning neural network corresponding to each subset, a step of carryingout unsupervised training on the second deep learning neural network, and a step of carrying out supervised training on the second or third deep learning neural network. The input data variable set is decomposed into multiple subsets, a deep learning neural network is initialized for each subset, then a deep learning neural network is initialized with output layers of all deep learning neural networks as input layers, the data dimensions and computational complexity of the deep learning are reduced, and thus the effects of the deep learning neural networks are improved.

Description

technical field [0001] The present invention relates to a deep learning neural network method and system, in particular to a deep learning neural network decomposition and synthesis method and system. Background technique [0002] When the number of input data variables and output data variables that the existing deep learning neural network can handle is very large, it will lead to a sharp increase in computational complexity, resulting in difficulty in training and poor application effect of the trained deep learning neural network. It can be understood that the number of data variables is the dimension of the data, so the more the number of data variables, the higher the data dimension, and the complexity increases exponentially. For example, the existing deep learning neural network can generally only learn the correlation between data at two time points during training, and then predict the data at another time point based on the data at one time point. If the input da...

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): G06N3/04G06N3/08
CPCG06N3/08G06N3/088G06N3/045
Inventor 朱定局
Owner SUPERPOWER INNOVATION INTELLIGENT TECH DONGGUAN CO LTD
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
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