Examples Rosenbrock

There are many variations of this method. To illustrate the procedure, a variation developed by Rosenbrock will be discussed. It is one of the best optimization methods known8,7 when there is no experimental error. The method is also very useful for determining constants in kinetic and thermodynamic equations that are highly nonlinear. An example of this type of application is given in reference 9. 

Example 2.4.1 Minimization of the Rosenbrock function by the simplex method of Nelder and Mead... 

Both the development and the optimization of simplex methods are still continuing. Several functions have been designed to test the performance of the simplex algorithms, one example is the famous ROSENBROCK valley. Other test functions have been reported by ABERG and GUSTAVSSON [1982]. Most analytical applications of simplex optimization are found in atomic spectroscopy [SNEDDON, 1990] and chromatography [BERRIDGE, 1990],... 

For example, in the method of Rosenbrock [211, 212], starting from the initial estimate a(0) the first step consists of varying successively each parameter one after the other and so of determining an improved value a(,). New directions of searching are then defined by a rotation of the axes, one of the axes joining the initial and final points of the first step. This method has been proved to work efficiently in the case of an edge on the hypersurface S(a) versus a. 

The resulting system is called a set of differential-algebraic equations (DAE) and their solution is now a specialised field with its own texts [130, 286] and there is a package program, DASSL [441], for their solution. This can be of use in the present context, for example with the method of lines, which indeed often results in a DAE system. This is gone into in some detail in Chap. 9, in the context of Rosenbrock methods. 

Before moving on to real Rosenbrock methods, consider again (4.66). The left-hand side contains a term in fy if we are dealing with a system of odes, this is called the Jacobian of the system. It is often constant, evaluable in advance. It will be seen in Chap. 9 that unless the diffusion problem has nonlinear concentration terms (for example from higher-order homogeneous reactions), the Jacobian is constant. If not, it must be evaluated at every step. 

A simple example serves to illustrate the use of Rosenbrock, using the ode... 

There is an example program described in Appendix C, BPROS, applying Rosenbrock to the Birk-Perone system, in which we have both time-dependence and nonlinear equations. It is described in Chap. 2, pages 22-22. Equation (2.73), with boundary conditions, when semidiscretised using equal intervals in space, leads to the DAE system... 

An illustration of simulated annealing optimization is presented in Figures 1-5. In this 2-dimensional example the objective fimction is a modified Rosenbrock s function ... 

For example, let us consider the well-known Rosenbrock s function problem ... 

Fig. 9.2 illustrates the Rosenbrock method for a model with two parameters. Because the algorithm attempts searching along the axis of a valley of the objective function S, many evaluations of S can be omitted compared with, for example, the algorithm with univariate... 

The example program CHRONO CN HERM shows a Crank-Nicolson implementation and again, the 2(3) form is only marginally better than 2(2). Bieniasz showed how to use this scheme for extrapolation and for the Rosenbrock ROWDA3 scheme [10]. The reader is referred to that paper for details, where still higher-order forms are found. The paper makes it clear that extremely small errors can be achieved by using this method. 

Here some modules, proeedures and whole programs are described, that may be useful to the reader, as they have been to the authors. They are all in Fortran 90/95 and invoke a generally useful module, that will be used in all procedures and programs in the examples, and another module useful for programs using a Rosenbrock variant. The source texts (except for the two modules) are not all reproduced here, but can be downloaded from the web site http //extras.springer. com. 

This option is given as a Boolean expression. For example, if stiff =true, the equation is considered stiff and the default method value is set to rosenbrock instead of rk45. ... 

---