Retreival Parameters

Fitting

TauREx 3 employs dynamic discovery of retrieval parameters. When a new profile/chemistry etc is loaded into a forward model, they also advertise the parameters that can be retrieved. The forward model will collect and provide them to the optimizer which it then uses to perform the sampling.

Classes that inherit from Fittable are capable of having retrieval parameters. Classes that inherit from this include:

• TemperatureProfile

• Chemistry

• Gas

• PressureProfile

• Star

• Planet

• ForwardModel

• Contribution

There are two ways of defining, fitting parameters. The simpler decorator method and programmaticaly

Decorator form

The decorator fitparam() decorator acts and behaves almost identically to the @property python decorator.

Programmaticaly

The decorator form is useful for describing parameters that always exist in a model in the same form. There are cases where parameters may need to be generated based on some input. One example is the NPoint temperature profile. Depending on the number of temperature points input by the user the temperature profile will actually generate new fitting parameters for each point (i.e T_point1, T_point2 etc) Another example are the Gas profiles. In this case, the fitting parameter names ar