Le Hoai Minh, Le Thi Hoai An, Pham Dinh Tao, Huynh Van Ngai: "An efficient DCA for spherical separation".

Abstract: The binary classification problem consists in finding a separating surface minimizing an appropriate measure of the classification error. Several mathematical programming-based approaches for this problem have been proposed. The aim of spherical seperation is to find, in the input space or in the feature space, a minimal volume sphere separating set A from set B (i.e. a sphere enclosing all points of A and no points of B). The problem can be cast into the DC programming framework. Afterwards, we propose a simple DCA scheme for solving the resulting DC program in which all computations are explicit. Computational results show the efficiency of the proposed algorithms over the two other spherical seperation methods: FC[6] and UCM[7].

Keywords: Classification, Spherical Seperation, DC Programming, DCA.

Citation: Le Hoai Minh, Le Thi Hoai An, Pham Dinh Tao and Huynh Van Ngai, An efficient DCA  for spherical separation, in “Intelligent Information and Database Systems”, Lecture Notes in Artificial Intelligence, volume 6592, pp 421-431, Springer Verlag 2011.