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

Sample contribution rate-based single detection credibility evaluation method for binary hard classifier

A technology of hard classifier and evaluation method, applied in the fields of instrument, calculation, character and pattern recognition, etc., can solve the problems of uneven distribution of training samples and low reliability, avoid high-risk decision-making, improve reliability, quantify effect of reliability

Pending Publication Date: 2020-11-13
HEILONGJIANG UNIV
View PDF0 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0008] The purpose of the present invention is to solve the problem that the reliability of a single detection result is determined only based on the probability of each interval due to the uneven distribution of training samples, and now provides a single binary classification hard classifier based on the sample contribution rate Evaluation Method of Reliability of Second Detection

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
  • Sample contribution rate-based single detection credibility evaluation method for binary hard classifier
  • Sample contribution rate-based single detection credibility evaluation method for binary hard classifier
  • Sample contribution rate-based single detection credibility evaluation method for binary hard classifier

Examples

Experimental program
Comparison scheme
Effect test

specific Embodiment approach 1

[0035] Specific implementation mode one: refer to figure 2 Specifically explaining this embodiment, a method for evaluating the reliability of a single detection of a binary hard classifier based on the sample contribution rate described in this embodiment includes the following steps:

[0036]Step 1: In the binary classification problem, when the hard classifier model is used for estimation, the samples in the set X are divided into positive samples and negative samples during the training process, and the corresponding vector distances are divided into positive distances and negative distances, that is Divided into category 1 and category 2 in the binary classification problem. For the training sample set X, the sample set X is input into the binary classification hard classifier model, and the distance d(X k ), k=1 , 2,...,Z, Z is the total number of samples in the sample set X.

[0037] Sample X k The distance from the hyperplane d(X k ) reflects the size of the samp...

specific Embodiment approach 2

[0054] Specific implementation mode two: refer to image 3 with Figure 4 This embodiment will be specifically described. In the sealed relay, the excess signal is an abnormal signal generated by the activated particles moving irregularly inside the sealed relay and colliding with certain internal structures. The component signal is generated because the movable component in the sealed relay collides with the inner wall of the cavity with the vibration of the vibrating table during the redundant object detection, and it is a normal signal generated. The one-time reliability evaluation method based on the sample contribution rate of the binary hard classifier is applied to the reliability evaluation of the component signal and the redundant signal of the sealed relay.

[0055] Table 1 shows the result of applying the distance between the sealed relay component signal and the excess signal sample and the hyperplane in this embodiment. Treating component signal samples as posi...

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 sample contribution rate-based single detection credibility evaluation method for a dichotomy hard classifier, and relates to the technical field of dichotomy detection credibility evaluation. The objective of the invention is to solve the problem of low credibility of determining a single detection result only according to the probability of each interval due to non-uniform distribution of training samples. The credibility evaluation method for single detection is designed by combining a credibility evaluation method of a binary classification hard classifier and integrating interval sample contribution rates aiming at the conditions of non-uniform distribution and large sample contribution rate difference of each interval section in actual sample distribution. High-risk decision is prevented in practical engineering application; reliability, accuracy and robustness of the system are improved.

Description

technical field [0001] The invention belongs to the technical field of binary classification detection reliability evaluation. Background technique [0002] In high-risk fields, such as the aerospace field, when testing the signal of sealed relays, in addition to caring about whether the prediction model can identify the category of the tested sample and the overall recognition accuracy, it is more concerned about whether the recognition result for a single sample is credible , that is: the reliability of a single recognition result. Take the detection of some precision instruments as an example. If the detection results are wrong, it will bring immeasurable losses, and more seriously, it will cause casualties. Therefore, for detection systems in the context of such high-risk fields, it is not only required to provide detection results, but also to provide reliability estimates of the output results, so that users can determine the processing method through the reliability ...

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): G06K9/62
CPCG06F18/2193G06F18/241
Inventor 王国涛赵广龙吕冰泽李响吴耀阳藤玉茹
Owner HEILONGJIANG UNIV