INVERSION OF FREDHOLM INTEGRAL-EQUATIONS OF THE 1ST KIND WITH FULLY CONNECTED NEURAL NETWORKS

Fredholm integral equations of the first kind are examples of ill-pose d problems in the Hadamard sense. Conventional methods of discretizati on often result in an ill-conditioned system of equations. The traditi onal approach is to solve this system using a variety of regularizatio n schemes. An alternative is to solve the system using connectionist p aradigms. The robustness of this approach is illustrated by using a fu lly connected network that resembles a Hopfield network.