A multi-instance dictionary learning and classification method and device based on similarity

Abstract

The invention discloses a similarity-based multi-example dictionary learning and classification method, comprising: obtaining a training set; for each positive candidate example selected from a positive bag set, repeatedly performing the following steps, cyclic iteration, training a classifier, Until the preset iteration stop condition is met: calculate the similarity weights of each example in the training set for the positive class and the negative class respectively; learn each example in the training set through dictionary learning to obtain the sparse coding of each example; learn a projection dictionary, The sparse coding of each example is re-represented; based on each example re-represented by projection and the similarity weight of each example, a classifier is trained; based on the trained classifier, the category of the target package is obtained. By applying the technical solution provided by the embodiment of the present invention, the category of the target package can be accurately determined, and the classification effect of the classifier is improved. The invention also discloses a multi-instance dictionary learning and classification device based on similarity, which has corresponding technical effects.

Description

technical field [0001] The invention relates to the field of computer application technology, in particular to a similarity-based multi-instance dictionary learning and classification method and device. Background technique [0002] With the rapid development of computer technology, multiple-instance learning has attracted more and more attention. Multiple-instance learning was proposed to solve the problem of classifying bags, which consist of many examples. For a positive bag, it contains at least one positive example, on the contrary, for a negative bag, it only consists of negative examples. [0003] Nowadays, multi-instance learning has been widely used in various real-world fields, such as: for drug molecule activity prediction, data mining, image classification, text classification, malware classification, etc. Taking image classification as an example, in most cases, the entire image is labeled instead of each region in the image. Such as figure 1 As shown, in th...

AI Technical Summary

Problems solved by technology

[0005] However, in practical applications, the fuzzy examples in the positive bag of the training set have a greater impact on the training of the classifier. The methods in the prior art do not take this influence into account, resulting in the unsatisfactory classification effect of the classifier. not tall

Images