1998 Volume 10 Issue 2 Pages 265-274
An idea of an optimization method for logic design is proposed. For an objective logic design, we construct a network that can demonstrate the logic circuit based on the neural computing. We call the network the logic neuron network. We train the network until it demonstrates the system behavior using the back propagation method. Next, we optimize the network and train it again. The above process is repeated until the network can't demonstrate the circuit. After that, we extract necessary features and parameters from the last converged network. Thus, we obtain an optimum solution by applying these processes to the objective circuit. In this paper, mainly we apply the method to the binary logic design. First, we define the AND, OR and EXOR logic neurons and show the correspondence between the neurons and logic gates. Next, we construct the objective logic networks using these logic neurons and show the optimization method. The logic neuron network is applied to binary logic design of Sum-of-Products, Product-of-Sums and EXOR Sum-of-Products expression. The simulation results shows a good solution. Thus, we prove that our proposed method is applicable to various logic designs.
