逻辑非兼平移CNN的鲁棒性设计

Robustness Designs of Logic Not Translation CNN

本文提出了一类细胞神经网络(CNN),这类CNN不仅能够将黑白图像中的黑白转换,同时还能将图像分别朝八个方向平移。鲁棒性分析给出了完成相应功能的参数不等式。数值模拟显示出理论结果在应用中的有效性。

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.