Robustness Designs of Logic Not Translation CNN
Xinjian Zhuo · Gongcheng shuxue xuebao · 2006
We design a set of cellular neural network (CNN) robustness templates. The CNNs can change each black pixel into white and vice-versa, and translate the changed images in 8 compass directions at the same time. The robustness analysis gives template parameter inequalities which guarantee the corresponding CNNs working well for performing prescribed tasks. Simulation examples are given.