Abstract:
The process of solving machine learning can be regarded as searching for a learning model with strong generalization ability and high robustness in the hypothesis space,and it is more difficult to find a suitable model in the hypothesis space.However,as a class of combinatorial optimization learning methods,ensemble learning can not only combine multiple simple models to obtain a better performance model,but also allow researchers to design combination schemes for specific machine learning problems to get a more powerful solution.This paper reviews the development history of ensemble learning,and focuses on the three strategies of diversity generation,model training and model combination in ensemble learning,and then describes the relevant application scenarios of ensemble learning at the current stage.Finally,the future research direction of ensemble learning is analyzed and forecasted.