Multi-model integration method and system

Abstract

The invention relates to the technical field of machine learning, and discloses a multi-model integration method, which comprises the following steps of: carrying out salient feature region extractionon each image in an original image data set to generate corresponding new images, and reconstructing all the new images into a plurality of feature image sets; performing model training on the original image data set to generate a corresponding original classification model; performing model training on each feature image set to generate a plurality of corresponding feature classification models;and integrating the original classification model and the plurality of feature classification models according to a preset model integration algorithm to obtain a final category prediction result. Correspondingly, the invention further discloses a multi-model integration system. According to the method, the influence of the homogeneous model on the integration method is reduced, and the overall accuracy of the model is improved.

Description

Technical field [0001] The present invention relates to the technical field of machine learning, in particular to a multi-model integration method and system. Background technique [0002] Ensemble learning is a type of learning algorithm in machine learning. It is a method of training multiple learners and combining them. When the ensemble method is applied to deep learning, a final prediction result can be obtained by combining the predictions of multiple neural networks. For example, the network snapshot integration method uses weighted snapshots to save several intermediate network parameters when training the model, and uses these models with the same storage efficiency but different weights to construct an integrated model after the training is completed. The integrated model can make the test set The classification performance has been improved. Under normal circumstances, integrating neural networks with different structures is a good method, because differentiated mode...

AI Technical Summary

Problems solved by technology

[0003] However, in the network snapshot integration method or some other integration methods, the same network structure is used to train the model on the same data set, although the saved basic model has different network parameters, but in a certain To some extent, it inevitably causes the basic model to have homogeneous properties, which reduces the benefits of the integration method brought by the differentiation between multiple models.

Images