Classification device, classification method and electronic equipment

Abstract

The invention provides a classification device, a classification method and electronic equipment for overcoming the problem that the classification score of a test sample acquired by utilizing a traditional graph-based learning method is not accurate. The classification device comprises: a clustering unit for clustering target samples; a determination unit for determining training samples related to each cluster of the target samples; a deleting unit for deleting classification scores of the training samples whose classification scores are not accurate; and a calculation unit for calculating the classification scores of the test samples according to the similarity between each test sample and each remaining training sample and the similarity between every two test samples by taking the target samples as the test samples. The classification method is used to execute the process capable of realizing the function of the classification device. The electronic equipment comprises the classification device. The technical scheme of the invention can be applied in the field of information processing.

Description

technical field [0001] The invention relates to the field of information processing, in particular to a classification device, a classification method and electronic equipment. Background technique [0002] As a method that can effectively describe the relationship between data, graph-based learning has been widely used in many fields, such as web page classification, image retrieval, video concept detection, etc. The above-mentioned webpage classification, image retrieval and video concept detection can all be regarded as a classification process in a broad sense. It should be noted that the graph mentioned here is a weighted graph, which is a data relationship, not an image in the true sense. [0003] Traditional graph-based learning methods usually use the similarity between training samples under a certain optimization framework to calculate an optimal solution for each training sample through an analytical expression or an iterative solution method. A category score r...

AI Technical Summary

Problems solved by technology

[0004] However, in the traditional graph-based learning methods mentioned above, the category scores of training samples often remain unchanged during the process of generalizing the learning results to test samples, which makes some training samples with inaccurate category scores likely to be negative The calculation of the class score of the test sample has a negative impact, that is, the calculated class score of the test sample is inaccurate

In addition, in order to generalize the learning results to test samples, the above-mentioned traditional graph-based learning methods usually process each test sample sequentially without considering the relationship between test samples, which may also make the category score of test samples inaccurate calculation

Images