No label defined (Q3790447)

From MaRDI portal

!

This is the item page for this Wikibase entity, intended for internal use and editing purposes. Please use the normal view instead:

scientific article; zbMATH DE number 4054781
Language Label Description Also known as
default for all languages
No label defined
    English
    No label defined
    scientific article; zbMATH DE number 4054781

      Statements

      0 references
      0 references
      1987
      0 references
      minimizing functions of several parameters
      0 references
      gradient methods
      0 references
      minimization of loss functions
      0 references
      software tools
      0 references
      nonlinear parameter estimation
      0 references
      nonlinear function minimization problems
      0 references
      solution of sets of nonlinear equations
      0 references
      interactive problem solving
      0 references
      BASIC
      0 references
      sensitivity
      0 references
      error distribution
      0 references
      Direct search methods
      0 references
      Hooke and Jeeves method
      0 references
      Nelder-Mead polytope method
      0 references
      conjugate gradient method
      0 references
      truncated Newton method
      0 references
      variable metric method
      0 references
      sums of squared functions
      0 references
      chemical kinetics
      0 references
      economic modeling
      0 references
      forecasting
      0 references
      Test problems
      0 references
      examples
      0 references
      This book provides researchers in many disciplines with an approach, via the minimization of loss functions, and with software tools to solve nonlinear parameter estimation (NLPE) problems using small computers. It is a book about mathematical software and about problem solving rather than mathematics. The authors have built a software system for NLPE related to the solution of nonlinear function minimization problems and to the solution of sets of nonlinear equations. The problems toward which this software is directed involve the optimization of mathematical functions in order to model or to improve the properties or fit to observed data of some model of a real system. The NLPE is carried out on a personal computer (PC), especially the IBM PC (or compatible), which allows interactive problem solving with easy access to program and data files. All programs are written in BASIC, since most microcomputers are equipped with a BASIC interpreter. The authors aim to provide programs operating in an interpretative environment. This is because the ability to quickly modify programs, print variables or array elements and, in some systems, change values and continue execution of a program is highly desirable in this particular context. A main goal of the text is to present the rudiments of the software ``glue'' to integrate the tools above into an easily used system for solving estimation problems. Extra programs or subprograms are included to aid the user in related tasks such as: to check program codes for the loss function or its derivatives; to incorporate bound constraints on parameters; to perform post estimation analysis to determine the sensitivity of the parameters. Both the approach and software are documented in a unified and consistent way. The statistical approach to estimation is described as follows: a) specify the mathematical form of the model of the system under study; b) specify the nature of the fluctuations in this model (error distribution); c) derive the relationship between the observed data and the unknown parameters of the model; d) use this relationship to determine the most probable parameter values or those which minimize the deviation between predicted and observed behaviour of the system. The NLPE cannot be fully automated because of the important decisions which must be made concerning model and error distribution forms, the rejection of outliers and the analysis of results. The authors point out the important details of coding of parameter estimation programs and of the loss functions and their derivatives, along with the background reasons for these details, which make a great deal of difference to the performance and success of programs and their users. The first three chapters are an informal introduction to the field. Their titles are: 1. Overview and purpose; 2. Examples of NLPE problems; 3. Exploring the problem. The next eight chapters deal with some methods for finding minima of functions: 4. Direct search methods; 5. The Hooke and Jeeves method; 6. The Nelder-Mead polytope method; 7. Minimization using gradients; 8. The conjugate gradient method; 9. The truncated Newton method; 10. The variable metric method; 11. Methods for sums of squared functions. The heart of the book is to be found in Chapters 12, 13, 14 and 15: Choosing an approach and method; Measures and consequences of nonlinearity; Difficult problems; Implementation details. The last three Chapters are: 16. Examples from chemical kinetics; 17. Applications in economic modeling and forecasting; 18. Test problems and other examples. The programs themselves, an extended form of bibliography, and more examples are contained on three 5.25'' disks (the last two are available from the second author). The book is photographically reproduced from typescript; in the age of \(T_ EX\) this is not satisfactory. The style is refreshing, and conveys a sense of enthusiasm and discovery making it a pleasure to read. I have considerable respect for the hard work and mastery that went into this book. There is a good mixture of explanation and demonstration of the techniques in action. One needs not to study the entire book to profit from it.
      0 references

      Identifiers

      0 references
      0 references
      0 references
      0 references
      0 references
      0 references