Mean iteration-based multi-task learning model training and prediction method

A multi-task learning and model training technology, which is applied in the field of multi-task learning model training and prediction, can solve the problems of insufficient utilization of data information and low operating efficiency of multi-task learning methods

Active Publication Date: 2020-01-31
NAT UNIV OF DEFENSE TECH
View PDF5 Cites 2 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0007] Aiming at the defects of low operating efficiency and insufficient utilization of data information in the prior art multi-task l

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
  • Mean iteration-based multi-task learning model training and prediction method
  • Mean iteration-based multi-task learning model training and prediction method
  • Mean iteration-based multi-task learning model training and prediction method

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0054] A multi-task learning model training method based on mean iteration includes the following steps:

[0055] (1) Obtain sample data sets of multiple tasks, and divide the sample data sets of each task into training set and test set.

[0056] There are T tasks, and for each task t (t=1, 2, 3...T), multiple instances containing D characteristic variables are collected, and the class label corresponding to each instance of each task is obtained at the same time to obtain each task Sample data set. Among them: the class label corresponding to each instance is available {C 1 ,C 2 ,...,C k } Means that K means the total number of class labels of all instances. x 1 ,x 2 ,...,X D Represents the characteristic variables of each instance, the total number of characteristic variables is D.

[0057] The sample data set of each task is divided into training set and test set. For the sample data set of the t-th task, Train t Represents the training set corresponding to the t-th task, where...

Embodiment 2

[0095] Reference figure 1 , Is a flowchart of a prediction method based on a multi-task learning model, the method includes:

[0096] Based on the multi-task learning model training method based on mean iteration provided in Example 1, a trained multi-task learning model is obtained.

[0097] For the instance n to be predicted in the tth (t=1, 2, 3...T) task among the T tasks, obtain the D feature variables x of this instance 1,n ,x 2,n ,...,X D,n , Based on the highest priority test values ​​of various labels on the t (t = 1, 2, 3...T) task finally output by the trained multi-task learning model And the conditional probability P of each feature variable on various labels in the tth (t=1, 2, 3...T) task t (x d |C k ), according to formula (11), the class label of the instance n can be obtained.

[0098]

Embodiment 3

[0100] A training method of student performance prediction model with multiple data sets, including

[0101] (1) There are multiple schools (one school corresponds to one task), and multiple instances with D characteristic variables are collected for each school. An example is a student. The corresponding D characteristic variables of the student can include the year of the exam, the percentage of students who are eligible for free school meals, the percentage of students at the first level of VR (the highest level of the oral reasoning test), and the gender of the school (S.GN.) , School denomination, student gender, student ethnicity, VR band (can be 1, 2 or 3). At the same time, get the class label corresponding to each instance of each task, and get the sample data set of each task. Divide the sample data set of each task t into training set and test set;

[0102] (2) Train a multi-task learning model;

[0103] (2.1) For each task, use its corresponding training set to obtain ...

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

A mean iteration-based multi-task learning model training and prediction method comprises the steps of firstly obtaining sample data sets of a plurality of tasks, and dividing the sample data sets ofthe tasks into a training set and a test set; for each task, utilizing the corresponding training set to obtain the prior probability of each class label on each task and the conditional probability of each characteristic variable in each task on each class label, and calculating the class label of each instance in the test set corresponding to each task; updating the prior probability of each type of label on each task based on a mean iteration method; and performing loop iteration continuously until the sum delta of absolute values of prior probability errors of class labels on different tasks is smaller than a set threshold, and performing convergence to obtain a trained multi-task learning model. According to the mean iteration-based multi-task learning model training and prediction method, the multi-task learning efficiency can be significantly improved. Meanwhile, shared information between tasks and priori knowledge of data can be more fully utilized, and a better classificationeffect can be achieved by using fewer computing resources.

Description

Technical field [0001] The present invention relates to the technical field of big data processing, in particular to a multi-task learning model training and prediction method. Background technique [0002] With the development of information technology and the advent of the era of big data, machine learning has become one of the important methods to solve practical problems. At present, most machine learning methods adopt single-task learning (Single-task Learning, STL) method, that is, the learning process between multiple tasks is independent of each other, this method ignores the correlation between tasks. When single-task learning methods solve complex problems, they often decompose the problem into simple and independent sub-problems, and then merge the results to obtain the solution of the complex problem. [0003] However, this seems reasonable, but it is actually incorrect. Because many problems in the real world cannot be simply decomposed into independent sub-problems,...

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
CPCG06F18/214
Inventor 周鋆孙立健符鹏涛朱先强张维明
Owner NAT UNIV OF DEFENSE 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