Unlock instant, AI-driven research and patent intelligence for your innovation.

Method and system for comparing samples trained by neural network

A technology of neural network training and convolutional neural network, which is applied in the sampling field of comparative neural network training, can solve problems such as the decline of classification effect and sample imbalance, and achieve the effect of enhancing diversity, improving model accuracy, and improving expression ability

Active Publication Date: 2019-05-21
GUILIN UNIV OF ELECTRONIC TECH
View PDF5 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

However, a verification loss function like this has a serious sample imbalance problem. Positive samples are usually much less than negative samples, resulting in a decline in the overall classification effect.

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
  • Method and system for comparing samples trained by neural network
  • Method and system for comparing samples trained by neural network
  • Method and system for comparing samples trained by neural network

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0042] The principles and features of the present invention are described below in conjunction with the accompanying drawings, and the examples given are only used to explain the present invention, and are not intended to limit the scope of the present invention.

[0043] Such as figure 1 As shown, the method for comparing neural network training sampling includes:

[0044] Step 1: collect data samples;

[0045] The data samples include binary classes.

[0046] Step 2: Build a convolutional neural network, and build a comparative information output layer and a comparative information loss layer in the output sequence of the fully connected layer;

[0047] Step 3: Computing the data samples through a comparison algorithm at the comparison information output layer to obtain comparison information and comparison labels;

[0048] The comparison algorithm includes:

[0049] output I (X i ,X j ) = Process_Function(X i ,X j )

[0050]

[0051] Among them, i, j represent ...

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 method and a system for comparing samples trained by a neural network. The method comprises the following steps: collecting a data sample; establishing a convolutional neuralnetwork, and establishing a comparison information output layer and a comparison information loss layer in an output sequence of the full connection layer; calculating a data sample on the comparisoninformation output layer through a comparison algorithm to obtain comparison information and a comparison label; resampling the comparison information and the comparison label to obtain a resamplingresult; performing loss value calculation on the resampling result in the comparison information loss layer through a loss algorithm to obtain a loss value during training; and adjusting the parameters of the convolutional neural network according to the loss value, so that the convolutional neural network learns the expression characteristics of the data sample to complete the training of the convolutional neural network.

Description

technical field [0001] The invention belongs to the technical field of large-scale image retrieval, and in particular relates to a sampling method and system for comparing neural network training. Background technique [0002] Sampling is an important step in supervised learning algorithms. When the sampling is unbalanced, the information learned by the algorithm will be biased towards the side with more samples, resulting in a decline in the overall recognition effect. [0003] Generally, there are two ways to deal with the problem of sample imbalance: oversampling and undersampling. Oversampling is usually to randomly resample the side with a small number of samples or perform sample expansion methods such as rotation, blurring, and noise to increase the number of samples. Undersampling is usually to randomly sample the party with a large number of samples to obtain a new subset of training samples to reduce the number of samples. [0004] Neural networks can achieve go...

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): G06N3/04G06N3/08
Inventor 蔡晓东曹艺
Owner GUILIN UNIV OF ELECTRONIC TECH