This paper proposes a face recognition method which is based on a Generalized Probabilistic Descent (GPD) learning rule with a three-layer feedforward network. This method aims to recognize faces in a loosely controlled surveillance environment, which allows (1) large face image rotation (on and out of image plane), (2) different backgrounds, and (3) different illumination. Besides, a novel light compensation approach is designed to compensate the gray-level differences resulted from different lighting conditions. Experiments for three kinds of classifiers (LVQ2, BP, and GPD) have been performed on a ITRI face database. GPD with the proposed light compensation approach displays the best recognition accuracy among all possible combination.