Constructing Support Vector Classifier Depending On The Golden Support Vector

TL;DR: This paper is devoted to present a durable algorithm to construct the well-known Support Vector Classifier, by capturing a unique Vector from each class of training instances, called the Golden Support Vector.

Abstract: In order to increase the processing speed of online learning applications represented in its exigent requirement of reducing the amount of Support Vectors, this paper is devoted to present a durable algorithm to construct the well-known Support Vector Classifier, by capturing a unique Vector from each class of training instances. We called that Vector the (Golden Support Vector). Our algorithm had adopted basic mathematical tools for its constructional phases. The algorithm starts with applying its Hybrid Enclosing Mechanism in order to enclose two sets of mapped instances (Vectors) with the most optimistic non-overlapped curved spaces analogous to the instances distribution. This mechanism is considered as a spring point that leads us directly to separate these spaces with a Strong Separating Hyperplane. From both sides of that Hyperplane, two parallel Supporting Hyperplanes will be released settling down on the first detected Vector, which we called the Golden Support Vector. Each Supporting Hyperplane with its acquired Golden Vector is considered to be the basis to construct the edges of Maximal Margin of Separation Space; which offers best generalization ability not only to the trained instances but also to guarantee high predictive test accuracy for future instances from the same distribution. Finally, the Optimal Separating Hyperplane will intermediate the space of that margin vanishing all other Vectors. Rare inescapable cases have been discussed, provided with modest solutions as suggestions for future works. Affluent pictorial figures have been spread to emphasize our algorithm credibility.