A New Non-Autonomous Chaotic Neural Network for Solving TSP.
Kushan Ahmadian, Abbas Golestani, Mohammad Reza Jahed‐Motlagh, Nasser Mozayani · International Conference on Artificial Intelligence · 2008
In this paper we propose a new chaotic noise injection strategy into the Hopfield network which belongs to a general class of chaotic neural networks called nonautonomous network. The objective of this method is to inject noise into the neurons based on the history of dynamics of their neighbor neurons. The measure of adjacency which is a criterion for definition of neighborhood, is based on the coexisting of neurons in a same cluster and those clusters are indicated by an ISODATA algorithm. Experimental results show efficiency of the proposed algorithm, especially in maps with separated clusters.