Autoassociative Local Neural Networks for Face Recognition

Using autoassociativity principle, local connections, weight sharing, and proximity of input pixels, a neural network is designed which exhibits strong face recognition properties. Local connections can be found by an exhaustive search but it appears that a random choice gives good results too. I. Introduction Pattern recognition and pattern autoassociation are related but not identical tasks attributed to intelligent systems. In pattern recognition, a system is supposed to identify a class C i , i = 1; : : : ; K, where an object belongs to, giving the object's features x which were previously measured and delivered to the system. In autoassociation, an associative memory reconstructs the original pattern x when distorted or incomplete version x 0 is presented to the system [1, 6, 2, 3]. Usually such a system is trained to keep many original patterns which can be considered as representatives of pattern classes. In a sense, pattern recognition is more hard activity than autoassocia..