Adaptable linear support vector machine

Linear Support Vector Machine (LSVM) has recently become one of the most prominent learning methods for solving classification and regression problems because of its applications in text classification, word-sense disambiguation, and drug design. However LSVM and its variations cannot adapt accordingly to a dynamic dataset nor learn in online mode. In this paper, we introduce an Adaptable Linear Support Vector Machine (ALSVM) which linearly scales with the size of training set. The most brilliant feature of ALSVM is that its decision boundary is adapted in a close form when adding or removing data.