| Author | Chien, Chiu-jung |
| Call Number | AIT Thesis no. CS-92-1 |
| Subject(s) | Computer-assisted instruction
|
| Note | A thesis submitted in partial fulfillment of the requirements
for the degree of Master of Science, School of Engineering and Technology |
| Publisher | Asian Institute of Technology |
| Abstract | A CAI Package for nonlinear least squares methods (DfMin)
provides four methods simplex, DSSRSSR, Quasi-Newton and
iterative, for nonlinear functions data fitting all using least
squares methods; and three methods -- simplex, DSSRSSR and Quasi-Newton, for unconstrained nonlinear functions minimization .
The user can get simple description of the algorithms involved
and detailed usage instructions before they actually run the
program and get enough information while he/she is asked to input
anything. Above all, the user can input function expressions from
keyboard or files directly .
The Simplex Method distorts simplex step by step, each step
closer to the minimizer using very simple strategy. It is
important especially while the quasi-Newton fails.
The DSSRSSR Method is a modified form of Direct Search with
Systematic Reduction of the Size of Search Region. It may need most
function evaluations, but provides a faithful hardworking minimizer
founder.
The Quasi-Newton Method is the Newton method combining a
globally convergent strategy -- line search, using finite differencing to calculate the gradient and positive definite secant update
(BFGS) factored form to calculate the Hessian. Normally, it runs
much quicker than the simplex and DSSRSSR especially for functions
expensive to evaluate.
Th~ Iterative Method minimizes the linear estimator residuals
using an iterative, least-squares technique directly, while the
other three methods treat the data fitting as problems of minimization of the sum of squared errors. Most of the time, it is the
quickest method to find the proper parameters .
Since no method is best to every case and no way to guarantee
the answer is a real minimum (except some special cases), it is
beneficial to collect several methods together, so that the user
can get answers from different methods to figure out better ensured results. |
| Year | 1992 |
| Type | Thesis |
| School | School of Engineering and Technology (SET) |
| Department | Department of Information and Communications Technologies (DICT) |
| Academic Program/FoS | Computer Science (CS) |
| Chairperson(s) | Huynh, Ngoc Phien |
| Examination Committee(s) | Phan Minh Dung ;Bohez, Erik L.J.
|
| Scholarship Donor(s) | Government of R. 0. C. ; |
| Degree | Thesis (M.Sc.) - Asian Institute of Technology, 1992 |