Method and device for constructing clustering model, medium and program product

Abstract

According to the embodiment of the invention, the invention provides a method and equipment for constructing a clustering model, a medium and a program product. The method described herein includes: determining a plurality of groups of augmented samples based on augmentation of a group of training samples, each group of augmented samples having the same category as the group of training samples; processing the plurality of groups of augmented samples by using a clustering model to determine a plurality of groups of clustering probability distributions associated with the plurality of groups of augmented samples, one group of clustering probability distributions indicating probabilities that each augmented sample in one group of augmented samples belongs to a plurality of predetermined categories; and updating parameters of the clustering model based on distribution differences among the multiple groups of clustering probability distributions. According to the embodiment of the invention, the clustering model can be trained in an unsupervised manner based on the difference between the clustering probability distributions of the augmented samples, thereby getting rid of the dependence on the labeled data.

Description

technical field [0001] Various implementations of the present disclosure relate to the computer field, and more specifically, relate to a method for constructing a clustering model, an electronic device, a computer storage medium, and a computing program product. Background technique [0002] With the development of computer technology, machine learning technology has been widely used in various fields. In the process of training a machine learning model, it is usually necessary to use a labeled training data set, which is also called supervised training. However, supervised datasets with annotations are often difficult to obtain due to the labor or time cost of the annotation process. Therefore, people expect to reduce the dependence on labeled data in the process of training machine learning models. Contents of the invention [0003] In a first aspect of the present disclosure, a method for constructing a clustering model is provided. The method includes: based on the...

AI Technical Summary

Problems solved by technology

However, supervised datasets with annotations are often difficult to obtain due to the labor or time cost of the annotation process.

Images