80 Convergence Analysis for Adding Decaying Self-Feedback Continuous Hopfield Neural Network

Because adding decaying self-feedback continuous Hopfield neural network (ADSCHNN) is proposed based on continuous Hopfield neural network (CHNN), firstly the CHNN is simplified after sigmoid activation function is replaced with piecewise linear activation function. Secondly, the extra self-feedbacks are added to the simplified CHNN to form simplified ADSCHNN. The convergence analysis is given for the simplified ADSCHNN. Finally it is proofed that ADSCHNN is more effective than CHNN, when they are applied to solve optimization problem and when ADSCHNN is applied to solve traveling salesman problem (TSP), the ADSCHNN with $zi<0$ is better than that with $zi>0$ and the energy of the ADSCHNN may increase.