Support Vector Machine: Feature selection with sequentialfs()

I want to programme a support vector machine. I have 60 features. Using the function sequentialfs() it goes either forward or backward through the matrix. The problem is that if I limit the number of features to n, it will stop after having found n features moving forward or backward through the matrix. However, then it maybe stops before it has found the optimal ones, because it never goes through the whole matrix. How can I assure that it finds the best combination and not only chooses them in order?