We regard a magnetic inverse problem as a spatial optimum allocation problem of currents and use a Hopfield neural network for solving this optimization problem. Because the Hopfield neural network has an initial state problem that the optimum solution cannot be obtained unless an initial state of network is set up suitably, adoption of a genetic algorithm is proposed for solving this initial state problem of Hopfield neural network. The effectiveness of proposed method is confirmed by computational simulations. ©1996 American Institute of Physics.