BP (Back-Propagation) neural network weight determination method and system and prediction method and system

A technology of BP neural network and determination method, applied in the field of computer performance management, can solve the problems such as the suspension of the adjustment process of the connection weight coefficient and the small derivative of the activation function.

Inactive Publication Date: 2018-07-17
HUBEI UNIV OF TECH
View PDF0 Cites 2 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

If the initial value of the weight of the neural network is too large, the input value after weighted summation will fall in the saturation area of ​​the activation function, resulting in a very small derivative of the activation function, and the correction formula for calculating the weight will lead to the connection weight coefficient the adjustment process will almost come to a standstill

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
  • BP (Back-Propagation) neural network weight determination method and system and prediction method and system
  • BP (Back-Propagation) neural network weight determination method and system and prediction method and system
  • BP (Back-Propagation) neural network weight determination method and system and prediction method and system

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0093] figure 1 It is a flow chart of the method for determining the weights of the BP neural network provided in Embodiment 1 of the present invention. Such as figure 1 Shown, a kind of determination method of BP neural network weight, described determination method comprises:

[0094] Step 101: Obtain a training sample set, where the training sample set includes several training samples.

[0095] Step 102: Obtain a plurality of BP neural network weights according to the correlation of each training sample.

[0096] Step 103: Construct a water wave group including several multi-dimensional water waves, and initialize the position, wave height and wavelength of each multi-dimensional water wave, wherein each position of the multi-dimensional water wave corresponds to a weight of the BP neural network.

[0097] The water wave feature selection rule encoding uses a water wave to represent a candidate solution for feature selection. The candidate solution set corresponds to a ...

Embodiment 2

[0156] figure 2 The structural block diagram of the system for determining the weight of the BP neural network provided by Embodiment 2 of the present invention. Such as figure 2 Described, a kind of determining system of BP neural network weight, it is characterized in that, described determining system comprises:

[0157] The training set obtaining module 201 is used to obtain a training sample set, which includes several training samples;

[0158] BP neural network weight acquisition module 202, for obtaining a plurality of BP neural network weights according to the correlation of each of the training samples;

[0159] The water wave group initialization module 203 is used to construct a water wave group comprising several multi-dimensional water waves, and randomly initialize the position, wave height and wavelength of each of the multi-dimensional water waves, wherein the position of each of the multi-dimensional water waves corresponds to one of the BP neural network...

Embodiment 3

[0189] A performance prediction method for optimizing BP neural network server, said performance prediction method comprising:

[0190] Step 31: Obtain the BP neural network weight of the training sample set, the BP neural network weight of the training sample set is the BP neural network weight determined according to the determination method described in any one of claims 1-4;

[0191] Step 32: training a classifier according to the weights of the BP neural network to obtain a trained classifier;

[0192] Step 33: Obtain the weight of the BP neural network to be tested for the sample to be classified, and the weight of the BP neural network to be tested is the weight of the BP neural network determined according to the determination method described in any one of claims 1-4;

[0193] Step 34: input the BP neural network weight to be tested into the trained classifier, and the trained classifier completes the classification of the BP neural network weight to be tested;

[01...

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 BP (Back-Propagation) neural network weight determination method, a BP neural network weight determination system, a prediction method and a prediction system. The determination method comprises the steps of acquiring a training sample set and a plurality of neural network weights; constructing a water wave group and initializing the water wave group; selecting a currentoptimal water wave and judging whether a fitness value is smaller than a fitness threshold value; if not, determining an optimal text characteristic vector; if not, performing propagation processing and calculating the fitness value; judging whether the fitness value of the water wave increases; if so, replacing the water wave before propagation processing with the water wave after propagation processing; updating the water wave group and iterative times according to each of second judgment results, and re-determining the current optimal water wave; judging whether the iterative times is smaller than an iterative threshold value; if so, judging whether the fitness value of the current optimal water wave reaches the fitness threshold value; and if not, determining an optimal neural networkweight. With the method and the system provided by the invention, the reliability of the neural network weight can be improved on the premise of guaranteeing the precision of classification.

Description

technical field [0001] The invention relates to the technical field of computer performance management, in particular to a BP neural network weight determination method, system and prediction method and system. Background technique [0002] The weight of the neurons of the BP neural network will affect the convergence of the objective function and its convergence speed. This is because the BP neural network system studies nonlinear problems, so the weight of the neural network will directly affect whether the error function can converge, whether it converges to a local minimum after convergence, and affects the time for training to complete. If the initial value of the weight of the neural network is too large, the input value after the weighted summation will fall in the saturation area of ​​the activation function, resulting in a very small derivative of the activation function, and the correction formula for calculating the weight will lead to the connection weight coeffi...

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): G06N3/08
CPCG06N3/084G06N3/086
Inventor 王春枝任紫扉叶志伟陈颖哲金灿王毅超吴盼周方禹王鑫孙爽蔡文成罗启星
Owner HUBEI UNIV OF 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