A new input-output function for binary hopfield neural networks
We present a new input-output function of the binary Hopfield neural network operating in a sequential mode and its application for solving combinatorial optimization problems. From convergence theorem for the binary network, we obtain the correct input-output function that satisfies the convergence conditions for any value of the self-conections. We also present performance comparison of different input-output functions through the N-queens problem. Our simulation results show that with our input-output function the network always reaches the global minimum in this problem. However, with McCulloch-Pitts or hysteresis McCulloch-Pitts functions the network is trapped in local minima.
Unable to display preview. Download preview PDF.
- Y. Takefuji, “Neural Network Parallel computing”, Kluwer Academic Publishers, 1992.Google Scholar
- M. Ohta, A. Ogihara, K. Fukumaga, “Binary neural network with self-feedback and its application to N-queens problem”, IEICE Trans. inf. & syst., vol. E77-D, no. 4, 1994, pp. 459–465.Google Scholar
- C. K. Cheng, T. Lee, “On the convergence of Neural Network for higher order programming”, Proc. of International Joint Conference on Neural Networks, 1993, pp. 1507–1511.Google Scholar