Rational function models

We discuss in this section several models used in optimizations. Of these, the most successful are the quadratic model and its modifications, the restricted second-order model and the rational function model. 

Figure 5 Solutions for the step-length parameter fi for the rational-function model for the function + lOy at (—15,10) and with S = 1/150.

Rational function models inherit the advantages of the polynomial family, despite a less simple form, and can take on an extremely wide range of shapes. They have better interpolation properties (typically smoother and less oscillatory) than polynomial models, and excellent extrapolation powers due to their asymptotic properties. Moreover, they can be used to model a complicated structure to a fairly low degree in both the numerator and denominator. On the other hand, because the properties of the rational function family are often not well understood, one might wonder which numerator and denominator degrees should be chosen. Unconstrained rational function fitting may also lead to undesired vertical asymptotes due to roots in the denominator polynomial. 

It is important to realize that whenever qualitative or frontier molecular orbital theory is invoked, the description is within the orbital (Hartree-Fock or Density Functional) model for the electronic wave function. In other words, rationalizing a trend in computational results by qualitative MO theory is only valid if the effect is present at the HF or DFT level. If the majority of the variation is due to electron correlation, an explanation in terms of interacting orbitals is not appropriate. 

We note the possible similarities of the effective model (42) with Golay s theory as presented in Paine et al. (1983). In the effective dispersion term this theory predicts a rational function of and we confirm it. Nevertheless, there is a difference in particular coefficients. 

There is another way to introduce restrictions on the step lengths in the global part of an optimization. The rational function (RF) model is given by9... 

Global strategies for minimization are needed whenever the current estimate of the minimizer is so far from x that the local model is not a good approximation to fix) in the neighborhood of x. Three methods are considered in this section the quadratic model with line search, trust region (restricted second-order) minimization and rational function (augmented Hessian) minimization. 

Development of chemical speciation schemes which can be directly related to measures of bioavailability - This would allow the determination of which active trace element species merit the most intensive research from the standpoint of environmental perturbation. Some studies have attempted to correlate metal fractions determined by a particular technique (operationally defined speciation) with those that are bioavailable (functionally defined speciation) (Larsen and Svensmark, 1991 Buckley, 1994 Deaver and Rodgers, 1996). However, any correlation is only empirical and more research is required to achieve an understanding of the mechanisms involved in bioavailability and to develop rational predictive models. 

Instead of artificially transforming the data to a linear model, our group developed an approach in which the relation between isotope ratios and mole ratios is described by means of a polynomial regression (Jonckheere et al., 1982). In this, the basic IDMS equation [Eq. (1)] is seen as a rational function ... 

Eqn. (3.42) is an example of a so-called rational function . Functions of this kind are renowned for their flexibility in describing curves without physical modelling. 

The obtained relationship for such /(co) turned out to be a rational function of the SF L(z). The obtained formula allows calculation of /(co) (for any specific molecular model) in all the frequency region, including the low-frequency part of the spectrum. In its high-frequency part such /(co) coincides with the Boitzmann susceptibility /B(co). 

Other model alternatives are higher order polynomials, rational functions of several variables, nonlinear PLS, neural networks, nonlinear SVM etc. With higher order polynomials, or with linearized rational functions, it advisable to use ridge regression, PLS, or some other constrained regression technique, see e.g. (Taavitsainen, 2010). These alternatives are useful typically in cases where the response is bounded in the experimental region see e.g. (Taavitsainen et. al., 2010). 

The full actuation model is represented by G s)H s). Since H s) involves non-rational functions, such as sinh(-), cosh(-), and a/, it is infinitedimensional. For practical implementation of feedback control design, however, finite-dimensional models are desirable. Simple model reduction steps can be taken to obtain finite-dimensional models for IPMC actuators, by exploiting the knowledge of physical parameters and specific properties of hyperboiic functions. In particular, based on the physical parameters of IPMCs (see Section 4.2.3), 7(s) 10, and K 10 , and we can make... 

An obvious defect of the Turing model is that, because of its linearity, it cannot describe saturation effects. To overcome this difficulty a two-variables model was introduced by Gierer Meinhardt (1972). Although its terms can be identified with reactions and diffusion, the equations contain rational functions at the right-hand side. Denoting by a and h the complete concentrations of the activator and inhibitor (and not the deviations from the equilibrium values) the model is... 

Although linearized polynomial relationships like those previously examined are commonly used in fitting procedures for chemical and physical data. King and Queen suggested that rational functions are surely more adequate for these aims [36]. So, other authors emphasized the use of this correlation procedure type, and the non-linear regression model for the representation of thermomechanical properties for some binary solvent systems has been elegantly discussed by Rolling in earlier works [37]. 

Irrespective of the situation, process system identification is focused on determining the values for Gp and G/ as accurately as possible. Since most of the applications assume that the controller is digital, the system identification methods considered here will focus on the discrete time implementation of system identification. For this reason, the models for each of the blocks will be assumed to be linear, rational functions of the backshift operator Such models are most often referred to as transfer functions. The most general plant model is the prediction error model, which has the following form ... 

Example 4.1. This example compares a discrete-time, rational transfer function model with the FSF model in terms of the effect of sampling rate on the model parameters. Consider a third order system with time delay given by... 

A rational function can also be considered as an empirical model. It is an algebraic fraction such that both the numerator and the denominator are polynomials ... 

---