Kernel-based online machine learning and support vector reduction

Experimental investigation and development of a SVM model for hydrogenation reaction of carbon monoxide in presence of Co–Mo/Al2O3 catalyst

“…Lagrangian multipliers ( i α ) can be found out by solving following quadratic programming problem [33][34][35][36][37] …”

Support Vector Number Reduction: Survey and Experimental Evaluations

“…The hardware-friendly SVM, aiming to implement the SVM on an embedded system, includes hardware-based online learning [40], feedforward SVM without multipliers using Laplacian kernel [41], and SVM with integer parameters [42]. This paper categorizes SVNR methods into five approaches by incorporating categories of previous works [29]- [32], [86] and analyzing them from the viewpoint of an application engineer. First, SVNR methods are divided into either prepruning or postpruning as in [29] according to whether they exploit the result of a standard SVM.…”

“…The methods in [25], [32], [33], and [65]- [69] search for a subset of the original SV set minimizing the approximation error. The method in [65] starts from an empty set and iteratively adds an SV that is expected to make the largest decrease in the approximation error.…”

“…It will be explained in detail in Section III-A. In [32] and [66], the approximation error, which generally requires a large number of calculation, is estimated with the span of SV. In [25], [33], and [69], a solution represents whether each of the original SV set will be included in the reduced SV set or not, and the optimal solution is searched for using either genetic algorithm or particle swarm optimization.…”

“…This paper categorizes SVNR methods that have been researched for a long period of time into five approaches and summarizes them. This categorization is proposed by incorporating categories of previous works [29]- [32], [86] and analyzing them from the viewpoint of an application engineer. Although SVNR has been researched for quite some time and has produced useful outcomes, it is little known to the field of recognition systems for IVs and ITSs.…”