Global minimization of a quadratic functional: neural network approach

N -dimensional vectors ) ,..., , ( 2 1 N s s s = s will be called configuration vectors. They define N 2 possible states of the system (the configurations), among which the optimal configuration with regard to the objective function ) (s E has to be found. The number of different states increases exponentially, when N increases. Already if N > 50, it is almost impossible to solve the problem by means of direct enumeration. In physical applications and in the theory of neural networks the matrix