An examination of least squares support vector machines (LS-SVMs) which are reformulations to standard SVMs. The authors explain the natural links between LS-SVM classifiers and kernel Fisher discriminant analysis. Bayesian inference of LS-SVM models is also discussed.