A convex method for selecting optimal Laguerre filter banks in system modelling and identification

When approximating dynamic systems with Laguerre Basis Functions it is important to tune the Laguerre pole such that the expansion is both parsimonious and accurate. The sum of squared error (SSE) objective function has many local minima and therefore cannot be optimized directly. Two alternative objective functions have been proposed in the literature: an asymptotically optimal objective function, and an enforced convergence criterion (ECC). Currently both objective functions can only be evaluated in a system modelling framework. Two questions that will be addressed in this paper are: (1) does minimizing the ECC lead to a good estimate of the optimal Laguerre pole position (in the SSE sense), and (2) is it possible to use the ECC in a system identification framework?