Class ModelParameters

source code

A model has a set of parameters associated with it. Each parameter has a name, units and a value. To display fit results there is an optional title separate from the name, and an expanded description such as might show up in a tooltip for a fit user interface.

The value may be constrained by natural bounds such as depth greater than zero or volume fraction between zero and one. The value may also be constrained to be a discrete integer such as the number of repeats in a layer model, though it is still represented by a double in the fit. More complicated constraints such as interface depth less than layer depth are not supported within the model parameter structure, though they may be possible in some fitting engines.

A set of parameters with the same name, bounds, etc., can be represented as a vector.

Within the model, start with a blank set of parameters and add them one at a time using:

from numpy import array from fit import ModelParameters P = ModelParameters() P.add('P1',array('d',[1]*5),tip='P1 is a vector of 5 elements') P.add('P2',4,lo=0,hi=1,tip='P2 is bounded by [0,1]') P.add('P3',5,discrete=True,tip='P3 is an integer defaulting to 5') P.add('P4',3,units='cm',tip='P4 is measured in centimeters')

The values of the parameters can be accessed by name from the model:

P.P1[3] = 4. P.P4 = 2

Note that declaring a parameter to be discrete doesn't force the value to be integer. In fact it will be stored as a float, and the model will have to force it to be an integer when it reads it. Not all fit engines will handle discrete values properly.

The values can also be accessed by number:

P[0] = 4. P[3] = 2

Other properties can also be accessed by number, such as title, tip and bounds.