Home > Publications > Parsing Parameter Estimation Problems from EASY-FIT to SOCS

Parsing Parameter Estimation Problems from EASY-FIT to SOCS

Abstract:

In many cases, mathematical models involve parameters that must be fit to experi- mental data. These so-called parameter estimation problems have many applications that may involve differential equations, optimization, and control theory. In this thesis we consider only parameter estimation problems that involve explicit model functions, ordinary differential equations, and differential-algebraic equations.

This thesis reviews two software packages, EASY-FIT and SOCS, which are used to solve parameter estimation problems [5], [3]. We discuss the design of a parser used to translate EASY-FIT input into SOCS input so that is it possible to quickly test SOCS on a number of parameter estimation problems varying both in size and difficulty.

After parsing a small subset of parameter estimation problems from each of the three categories given above, we find that the parser performs very well. We are able to test SOCS on this subset of problems in a matter of seconds. This is a small fraction of the time it would take to code each problem separately in SOCS. Although there were differences in some of the solutions found by EASY-FIT and SOCS, they do not appear to be a result of the parser.

Author: Matthew Donaldson

Advisor: Raymond J. Spiteri

Download: mdonaldson_bsc_thesis