A method for evaluating the learning performance of a machine learning system
A machine learning and performance technology, applied in machine learning, instruments, computing models, etc., can solve problems such as unsatisfactory evaluation times
Active Publication Date: 2018-07-24
SHANXI UNIV
View PDF4 Cites 0 Cited by
- Summary
- Abstract
- Description
- Claims
- Application Information
AI Technical Summary
Problems solved by technology
However, this method cannot meet the user's requirement of further increasing the number of evaluations m
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 moreImage
Smart Image Click on the blue labels to locate them in the text.
Smart ImageViewing Examples
Examples
Experimental program
Comparison scheme
Effect test
Embodiment Construction
[0043] The performance of machine learning systems is often characterized by generalization error. In theory, the generalization error is the mean value of the loss of a machine learning system over the data population. Since the data population is not available in practice, we can only use a data set with multiple records to estimate the generalization error. The accuracy of an estimate of generalization error is mainly determined by the deviation of the estimate from the true value and the variance of the estimate itself. A good estimate has less bias and less variance.
[0044] In order to accurately estimate the generalization error of the machine learning system, the user needs to divide the data set into multiple sets of training sets and verification sets through a specific data segmentation method. Currently, the m×2 cross-validation method is one of the commonly used data segmentation methods. This is mainly because the m×2 cross-validation method has a better effe...
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
Login to View More Abstract
The invention discloses a machine learning system learning performance evaluation method. The method comprises the following steps: according to evaluation frequency given by a user, cutting a data set into multiple different training sets and verification sets; for each training set and each verification set among the multiple training sets and the verification sets, training a machine learning system by use of the training set (the verification set) so as to obtain a machine learning model; then by use of the verification set (the training set), testing the machine learning model so as to obtain a single estimation of the performance of the machine learning system; when the multiple training sets and the verification sets are all used, averaging all the estimations of the performance of the machine learning system as a final estimation of the system performance; at the same time, waiting to see whether the user adopts the current estimation; if the user needs to increase the evaluation frequency, gradually increasing residual training sets and test sets on an original basis, and executing training and testing of the machine learning system until a new performance estimation is calculated; and if the user adopts the current estimation, returning current estimation of the performance of the machine learning system.
Description
technical field [0001] The invention relates to an evaluation technology of a machine learning system, in particular to a method for evaluating the learning ability of a machine learning system. Background technique [0002] Machine learning system is an important system for intelligent processing and analysis of data. It learns based on existing data sets and applies the learned model to future information predictions. With the advent of the data age, machine learning systems have sprung up in various automation scenarios. For example, spam filtering systems based on machine learning algorithms, sentence sentiment classification systems, etc. all belong to the category of machine learning systems. [0003] With the advent of the Internet era, a large number of machine learning systems have been developed accordingly. The performance of these machine learning systems varies. Therefore, system developers and users must use data sets containing multiple records to objective...
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
Login to View More Patent Type & Authority Patents(China)
IPC IPC(8): G06N99/00
CPCG06N20/00
Inventor 王瑞波
Owner SHANXI UNIV