Inequality (22) indicates that the maximum-norm is the loosest among all p-norms. Fortunately, this loosest constraint would not seriously affect the accuracy since the value of ||y||∞ is comparable to that of the 2-norm and 1-norm. The maximum-norm provides us with the largest number of possible solutions under a given error limitation [24]. This would greatly enhance the possibility of finding a group of optimized coefficients when scanning a vast solution set. On the other hand, checking the maximum deviation sounds more reasonable than checking the “distance” between the accurate and approximated wave numbers since it is not working in the space domain. Therefore, we chose the maximum-norm as our criterion for designing the objective functions to extend the accurate wave number coverage as widely as possible.
