In this paper, we are concerned with a class of<br /> neural networks with Mexican-hat-type activation functions. Due<br /> to the different structure from neural networks with saturated<br /> activation functions, a set of new sufficient conditions are presented<br /> to stu
