Premium
Multi‐innovation Newton recursive methods for solving the support vector machine regression problems
Author(s) -
Ma Hao,
Ding Feng,
Wang Yan
Publication year - 2021
Publication title -
international journal of robust and nonlinear control
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.361
H-Index - 106
eISSN - 1099-1239
pISSN - 1049-8923
DOI - 10.1002/rnc.5672
Subject(s) - support vector machine , computer science , identification (biology) , convergence (economics) , algorithm , newton's method , function (biology) , mathematical optimization , relevance vector machine , binary classification , mathematics , artificial intelligence , nonlinear system , botany , physics , quantum mechanics , evolutionary biology , economics , biology , economic growth
The support vector machine has been widely used in binary classification applications because of the simplicity of its implementation. This article proposes an online identification method based on the support vector machine in the field of parameter identification. By substituting the constraint item into the original criterion function to form a new criterion function about the weight vector and the bias item, and then on the basis of the recursive identification methods, the newly established criterion function is minimized by using the Newton search to derive the Newton recursive support vector machine algorithm. Additionally, in order to improve the performance of the Newton recursive support vector machine algorithm, the multi‐innovation identification theory is applied to optimize the algorithm, and a multi‐innovation Newton recursive support vector machine algorithm is derived. In addition, the convergence and the calculation amount of the proposed algorithms are carefully analyzed in this article. Finally, a numerical simulation example is given to compare the Newton recursive algorithm and the corresponding multi‐innovation recursive algorithm. The results show that the algorithms and optimization strategy studied in this article are effective.