Feature Selection For High Dimensional Data