Method for obtaining sparse solution of robust least square support vector machine

Abstract

The invention provides a method for obtaining a sparse solution of a robust least square support vector machine. The method is used for solving the technical problem of solution lack sparsity existing in an obtaining method of a solution of the robust least square support vector machine and comprises the implementation steps that training data is input; a robust least square support vector machine model is constructed; the robust least square support vector machine model is converted into a DC plan; the DC plan of the robust least square support vector machine model based on original space is constructed; the DC plan of the robust least square support vector machine model based on the original space is smoothed; a subset B and a matrix P are selected from a sample index set; the sparse solution of the smoothed DC plan of the robust least square support vector machine model based on the original space is obtained; whether an optimal solution is sought or not is judged. According to the method, the sparse solution of the robust least square support vector machine can be obtained, and the method can be applied to the fields of classification and regression of big data with noise.

Description

technical field [0001] The invention belongs to the technical field of data mining, and relates to a method for solving a solution of a robust least squares support vector machine, in particular to a method for solving a sparse solution of a robust least squares support vector machine, which can be applied to noisy environments Big data for classification and regression fields. Background technique [0002] Support vector machine (Support Vector Machine, SVM) has strong learning ability and generalization ability, and avoids the local minimum problem in traditional methods, so it is an important tool for data mining. However, the constraints of SVM are inequality constraints, and the solution is a quadratic programming problem, so the solution process requires a large amount of storage space. Least Squares Support Vector Machine (LSSVM) was proposed by Suykens in 1999. Compared with the standard SVM, LSSVM has many advantages. For example, the constraints of LSSVM are equa...

AI Technical Summary

Problems solved by technology

However, Wang et al.’s robust LSSVM solution based on dual theory and Yang’s algorithm both need to use all training samples to calculate the kernel matrix, and the obtained solutions do not have sparsity

In practical applications, this is undoubtedly very memory-consuming and slows down the training speed

Even, for some big data problems, due to computer memory limitations, these methods are difficult to implement on ordinary computers

Images