Criterion module
The criterion module allows the composition of several criteria to form the one you/the algorithm want.
Usual criteria
- IterationCriterion(iterations_max = 1000) stops the optimization when the number of iterations is reached
- MonotonyCriterion(ftol = 0.01) stops the optimization when the relative value of the function augmentes more than ftol
- RelativeValueCriterion(ftol = 0.01) stops the optimization when the relative change of the function value is below ftol
- AbsoluteValueCriterion(ftol = 0.01) stops the optimization when the absolute change of the function value is below ftol
- RelativeParametersCriterion(xtol = 0.01) stops the optimization when the relative change of the parameters is below xtol for each component
- AbsoluteParametersCriterion(xtol = 0.01) stops the optimization when the absolute change of the parameters is below xtol for each component
- GradientCriterion(gtol = 0.01) stops the optimization when the gradient is below gtol for each component
Composition of criteria
- AndComposition() composes several criteria and returns True when all the criteria return True as well
- OrComposition() composes several criteria and returns True when one or more of the criteria return True
The composition for usual criteria can be done with the criterion function. The composed criterion is an OrComposition with IterationCriterion, RelativeValueCriterion, RelativeParametersCriterion and GradientCriterion if iterations_max, ftol, xtol or gtol is given as a formal argument.
Information criteria
- AICCriterion
- ModifiedAICCriterion
