Parameter identification or model calibration,
as part of model building, deals with the computation of model unknowns (initial or boundary conditions and parameters) from experimental data.
Parameter identification is usually formulated as a non-linear optimization problem aimed to find the model unknowns which minimize some measure of the distance among model predictions and experimental data.
For the case of (large scale) non-linear models solving such a problem
is usually a very challenging task due mainly to the presence of several
suboptimal solutions or of several equivalent solutions, in other words,
to poor or lack of practical identifiability.
AMIGO is a multi-platform (Windows and Linux) toolbox which covers all the steps of the iterative identification procedure: local and global sensitivity analysis, local and global ranking of parameters, parameter estimation, identifiability analysis and optimal experimental design.
The ultimate goal is to enable the computation of model unknowns with the maximum accuracy and at a minimum experimental cost.
Follow the links for a
detailed problem definition and toolbox description.
- Maximum flexibility for the definition of models and observation functions.
- Multi-experiment tasks with local (experiment dependent) and global information.
- Multiple types of experimental noise conditions and, accordingly, different types of cost functions for parameter estimation and experimental design.
- Maximum flexibility for the definition of unknowns: parameters and initial conditions that may be local (experiment dependent) or global for all tasks.
- Several approaches to perform identifiability analyses: i) the use of the Fisher Information Matrix (FIM) to asymptotic analyses; ii) the plot of two-dimensional projections of the parameter estimation cost function and iii) the robust analysis by means of a Monte-Carlo based approach.
- Sequential-parallel optimal experimental designs formulated as general optimal control problems.
- A suite of state of the art numerical methods for simulation and optimization to cover a broad range of problems: integration of stiff, non-stiff and/or sparse dynamic systems, plus solvers for convex and multimodal nonlinear optimization problems.
- Enables an enhanced operating mode by automatically interfacing to FORTRAN.
- Generates reports and plots according to user requirements.
- Inputs via scripts or the Graphical User Interface.