Active learning method based on covariance matrix

Abstract

The invention discloses an active learning method based on a covariance matrix, and the method is characterized in that the method comprises the steps: collecting a butane concentration value in a debutanizer production process, and taking the butane concentration value as a soft measurement modeling sample; dividing the soft measurement modeling sample into a training set and a test set, whereinthe training set is divided into a labeled sample set and an unlabeled sample set; establishing a Gaussian process regression model by utilizing the labeled sample set, and determining initial parameters of the model; according to the label-free sample set, selecting a sample forming the maximum determinant value of the covariance matrix; re-establishing a Gaussian process regression model, and determining model parameters; and predicting the butane concentration by utilizing test set. According to the method, the effectiveness of the covariance matrix selection strategy is verified through numerical simulation analysis and application simulation of the debutanizer process.

Description

technical field [0001] The invention relates to the field of machine learning and is applied to the soft measurement of butane concentration in a debutanizer, in particular to an active learning method based on a covariance matrix. Background technique [0002] Traditional machine learning methods must rely on a large number of samples with complete information to build models, but in actual situations, due to artificial or environmental conditions, it is difficult to obtain complete sample information, and most of the sample information is missing; in this case , how to use a small number of labeled samples and a large number of unlabeled samples to improve model performance has become a key issue in machine learning research; if only a small number of labeled samples are used to train the model, the prediction accuracy and generalization ability of the model are difficult to achieve ideal results; In addition, ignoring a large number of unlabeled samples is a waste of data...

AI Technical Summary

Problems solved by technology

[0002] Traditional machine learning methods must rely on a large number of samples with complete information to build models, but in actual situations, due to artificial or environmental conditions, it is difficult to obtain complete sample information, and most of the sample information is missing; in this case , how to use a small number of labeled samples and a large number of unlabeled samples to improve model performance has become a key issue in machine learning research; if only a small number of labeled samples are used to train the model, the prediction accuracy and generalization ability of the model are difficult to achieve ideal results; In addition, ignoring a large number of unlabeled samples is a waste of data resources; therefore, it is necessary to label unlabeled samples. Commonly used algorithms include semi-supervised learning and active learning; semi-supervised learning aims to use the input and output modeling of labeled samples On the basis of extracting useful information in unlabeled samples to achieve the purpose of improving regression accuracy; however, although traditional semi-supervised learning methods improve model performance, they may increase the amount of calculation and improve the accuracy of the model to a large extent. depends on the structure of the semi-supervised model

Images