Multi-mark distance measurement learning method based on interactive modeling

Abstract

The invention discloses a multi-mark distance measurement learning method based on interactive modeling, and the method comprises the steps of extracting the training data of any multi-mark application scene, and carrying out the multi-mark marking on the training data; preprocessing the extracted training samples to improve the quality of the samples; expressing a distance measurement matrix to be learned as a combined distance measurement form; defining the multi-mark semantic similarity based on feature and mark collaborative calculation, and constructing a triple constraint set; constructing a multi-mark distance measurement learning model in combination with the combined distance measurement and the triple constraint set, and performing optimization solution on the multi-mark distancemeasurement learning model; after learning the distance measurement, mapping the training data to the distance measurement space, and then learning by using an existing multi-mark learning algorithmto obtain a multi-mark classifier based on the distance measurement learning; and inputting a to-be-predicted sample into the above classifier to obtain a labeled sample. According to the method, thetime complexity of the multi-mark learning system can be greatly reduced, and the practicability of the multi-mark learning framework is improved.

Description

technical field [0001] The invention relates to a multi-label distance metric learning method based on interactive modeling, in particular to a multi-label distance metric learning method based on interactive modeling of feature space and label space, which is applicable to any multi-label learning scene and belongs to machine learning technology field. Background technique [0002] In recent years, multi-label learning has received extensive attention from researchers and a large number of research results have emerged. However, due to the compositional nature of the label space, multi-label learning has high complexity and is difficult to apply to practical scenarios. The existing multi-label learning methods mostly start from the label space to model the correlation between labels, and do less processing of the feature space. Analyzing multi-label data, there is redundancy in the feature space. Therefore, how to construct a suitable multi-label distance metric represent...

AI Technical Summary

Problems solved by technology

However, due to the compositional nature of the label space, multi-label learning has high complexity and is difficult to apply to practical scenarios

Existing multi-label learning methods mostly start from the label space to model the correlation between labels, and do less processing of the feature space

Images