GPU-based modeling method for automatic machine learning

Abstract

The invention discloses a GPU-based modeling method for automatic machine learning. The method comprises the steps of establishing a model library; preparing data; conducting the normalization treatment on the feature part of the input data; selecting a plurality of models in the common mode to subject the models to the GPU parallel computation; automatically preparing data for cross verification according to the input data, wherein the default is 10 folds: 9 folds of training and 1 fold of verification test; respectively calculating the cross verification for the plurality of models through the GPU to obtain the errors of all the models respectively; selecting one model smallest in error as a finally selected model; adopting all data as training data; training the finally selected model to obtain a final model, and marking the final model as Ypre = f (x). In this way, an optimized model is automatically calculated and obtained for a common user. Therefore, the complexity of the machine learning application for the common user and the excessive dependence of the common user on experts are reduced.

Description

Technical field [0001] The present invention relates to a modeling method of machine learning, in particular to a modeling method of automatic machine learning based on GPU. Background technique [0002] Machine learning is a subject that studies how to use machines to simulate human learning activities. A more strict formulation is: machine learning is the study of machines to acquire new knowledge and new skills, and to recognize existing knowledge. [0003] Machine learning is a science of artificial intelligence. The main research object in this field is artificial intelligence, especially how to improve the performance of specific algorithms in empirical learning. Machine learning generally involves two major processes, one is the training process to obtain parameters or models, and the other is the application of models for prediction or classification. [0004] The training process of machine learning is complex, and it usually requires professionals to design a model and tr...

AI Technical Summary

Problems solved by technology

[0006] At present, the acquisition of machine learning models generally relies on machine learning experts to design some models, and then uses computer training to obtain optimized parameters and obtain the final model. This method is not only dependent on experts and cannot be widely used by ordinary users, but also subject to experts. Individual level and available computing resources

[0007] In order to improve the accuracy of the model, it is necessary to choose to expand the range of model selection and parameters, but this means a huge amount of calculation, which cannot be quickly and effectively provided by ordinary computing platforms, and needs to be completed through new methods

Images