Neural Network Strategies for Solving Synthesis Problems in Non-conflilct Cases in Distributed Expert Systems

In this paper, two neural network mechanisms for synthesis of solutions in non-connict cases in distributed expert systems (DESs) are proposed. The ideas are: inputs of the neural network are diierent solutions for the same problem from diierent expert systems in DESs; outputs of the neural network are the nal solutions for the problem after combining diierent solutions which should be the same as the human experts' nal solutions. The rst point is to set up the architecture of the neural network and train the neural network by adjusting weights of links to match the outputs of the neural network against the human experts' solutions for all patterns. The second point is that the neural network mechanism proposed in this paper can accommodate the variable number of inputs and outputs without changing neural network architecture.