Robust design of bipolar wave cellular neural network with applications

The robust design for cellular neural network (CNN) is an important issue, since no template parameters of CNN can be realised exactly in practice. Bipolar wave (BW) CNN is able to simulate the phenomena that black wave and white wave propagate on grey cells, collide and keep balance finally. Firstly, this paper establishes a theorem for designing the robust templates for BW CNN. The theorem provides a group of parameter inequalities to determine the template parameter intervals within which the templates can implement corresponding functions. Secondly, this paper sets up an optimal model for searching the template with maximum robustness for BW CNN. A set of optimal templates is obtained by using the method of brute-force search. Finally, one simulation example is provided to illustrate the effectiveness of the theorem and optimal model. The example also shows that the BW CNN can be used to segment two kinds of point sets in image.