Deep sample learning method based on iterative mean clustering

Abstract

The invention discloses a deep sample learning method based on iterative mean value clustering, which is carried out according to the following steps: S1: select training data, and obtain N+1 layer training sample subsets through N times of iterative mean value clustering algorithm processing, N≥ 1; S2: Regression training is performed independently on each layer of training sample subsets to obtain N+1 regressors; S3: Select verification data and send the verification data to N+1 regressors to obtain N+1 verification Results; S4: Based on the weighted fusion mechanism, determine the optimal weight corresponding to each regressor (w 0 ,w 1 ,...,w N ); S5: Obtain test data, and use N+1 regressors and corresponding optimal weights to obtain the final prediction result. The effect is: clustering the learning samples through multiple iterative mean values ​​to obtain different training sample data sets, and then performing training and learning respectively. In the case of the same number of samples, the learning ability of the model is effectively increased, and the classification or prediction is improved. accuracy.

Description

technical field [0001] The invention relates to artificial intelligence technology, in particular to a deep sample learning method based on iterative mean value clustering. Background technique [0002] With the development of artificial intelligence technology, there are various ways of sample learning, and the quality of sample learning methods seriously affects the accuracy of subsequent classification and regression. [0003] Most of the artificial intelligence algorithms in the prior art use a single sample data set for learning and training. On the one hand, due to the limited number of learning samples that can be directly obtained, the performance of the classifier or regressor is enhanced only by increasing the number of iterations. The effect is limited; on the other hand, the degree of authenticity of existing learning samples will also have a serious impact on the performance of the training model. If all learning samples are treated equally, it is difficult to a...

AI Technical Summary

Problems solved by technology

[0005] However, through the mechanism of online learning, it is necessary to gradually extract new learning samples, which increases the complexity of the algorithm, and the improvement of algorithm performance requires a relatively long process, and the initial performance is relatively poor

Images