Machine learning optimization method, system, computer storage medium and electronic device

An optimization method and machine learning technology, applied in the field of artificial intelligence, can solve problems such as insufficient training data, achieve targeted results, improve performance, and improve labeling efficiency

Inactive Publication Date: 2018-12-28
北京墨丘科技有限公司
View PDF0 Cites 2 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0003] However, in some application scenarios in reality, the training data is not very sufficient, which leads to the fact that the existing training data often has features that cannot be covered, that is, the collection of training data cannot ensure that all features are fully covered.

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
  • Machine learning optimization method, system, computer storage medium and electronic device
  • Machine learning optimization method, system, computer storage medium and electronic device
  • Machine learning optimization method, system, computer storage medium and electronic device

Examples

Experimental program
Comparison scheme
Effect test

example 1

[0093] still with figure 1 Take the convolutional neural network model shown as an example. Get a new picture (hereinafter referred to as the test picture) as input, and input it into the machine learning model that turns on the classification mode after training. The test picture has all the features of the third-layer network. The probability distribution results of the 20 categories obtained in the figure are as follows Image 6 as shown ( Image 6 Only shown with figure 2 distinguishing features image 3 .2, see other feature maps figure 2 ),based on Image 6 The probability distribution results shown in the calculation feature image 3 .1 and characteristics image 3 A standard deviation of .2:

[0094]feature image 3 The standard deviation of .1 is:

[0095] Std Dev(0.005,0,0.2,0,0.01,0,0,0.3,0,0.01,0,0.45,0,0,0.005,0.01,0,0,0.01,0)=0.12203

[0096] feature image 3 The standard deviation of .2 is:

[0097] Std Dev(0.06,0.05,0.04,0.06,0.05,0.05,0.04,0.06,0...

example 2

[0109] by figure 1 Taking the convolutional neural network model shown as an example, input the additionally obtained new test picture 3 into the network with classification mode turned on after training, the following can be obtained: Figure 8 The probability distribution of the feature map shown ( Figure 8 Only shown with figure 2 distinguishing features image 3 .2, see other feature maps figure 2 ):

[0110] Specifically, according to Figure 8 Features shown image 3 .2 Calculate the standard deviation of the probability distribution as follows:

[0111] StdDev(0,0.005,0.2,0,0.01,0,0,0.3,0.45,0,0.01,0,0,0,0.005,

[0112] 0.01,0,0,0.01,0) = 0.12203

[0113] Through the comparison of the above results and the threshold of recognition degree, it can be concluded that: for features with low recognition degree image 3 .2 has a high degree of recognition, so the new test picture 3 is considered to be sample data that can train the model to improve performance sign...

example 3

[0125] Assuming that the training data is 1000 trained pictures, the 1000 trained pictures are sequentially input into the machine learning model that starts the classification mode after training, that is figure 1 The machine learning model is shown, and the statistics of 1000 pictures correspond to each category obtained by all the feature maps extracted by each layer of the network in the first 5 layers of the machine learning model.

[0126] Taking the first training picture (hereinafter referred to as training picture 1), all feature maps (feature image 3 .1. Features image 3 .2. Features image 3 .3 and features image 3 .4) The obtained categories are as follows: figure 2 Shown:

[0127] feature image 3 .1 The categories obtained include: categories 1, 3, 5, 8, 10, 12, 15, 16 and 19;

[0128] feature image 3 .2 The categories obtained include: categories 2, 3, 10, 11, 14, 15, 16, 18 and 19;

[0129] feature image 3 .3 The acquired categories include: cate...

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 an optimization method, a system, a computer storage medium and an electronic device for machine learning, wherein, the obtaining method comprises the following steps of: obtaining at least one learning data; identifying the learning data to obtain an identification result, the identification result comprising: a feature map of the learning data and / or a category to which the learning data belongs; when the recognition result satisfies a preset selection rule, the learning data corresponding to the recognition result being used as the optimization data of the machine learning model. The invention can pertinently obtain the data which can obviously improve the performance of the machine learning model.

Description

technical field [0001] The invention belongs to the technical field of artificial intelligence, and in particular relates to a machine learning optimization method, system, computer storage medium and electronic equipment. Background technique [0002] Machine learning is an important branch of artificial intelligence. In recent years, due to the rapid development of artificial intelligence, machine learning also needs to continuously improve its learning ability. Machine learning refers to the ability to continuously improve its performance by continuously using new training data to learn and train it. For example, the performance of a neural network model can improve as the amount of training data increases. Therefore, based on a sufficient amount of training data, the neural network model can achieve expected performance improvements. This is because the training data is sufficient, so that most of the features of the data can be traversed during learning and training. ...

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): G06N99/00G06N3/04
CPCG06N3/045
Inventor 张昊孙鹏飞
Owner 北京墨丘科技有限公司
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