Face recognition has received extensive attention due to its potential applications in many fields. To effectively deal with this problem, a novel face recognition algorithm is proposed by using the optimal kernel minimax probability machine. The key idea of the algorithm is as follows: First, the discriminative facial features are extracted with local fisher discriminant analysis (LFDA). Then, the minimax probability machine (MPM) is extended to its nonlinear counterpart by using optimal data-adaptive kernel function. Finally, the face image is recognized by using the optimal kernel MPM classifier in the discriminative feature space. Experimental results on three face databases show that the proposed algorithm performs much better than traditional face recognition algorithms.