-
…ients argument (#1847) The eval_gradients() methods duplicate much of the code of the eval() methods. To increase the maintainability of the code, the eval_gradients() methods have been removed, and the computation of partial derivatives is now done by the eval() methods if the optional gradients argument is set to true.
-
The GModelSpectralRegistry::alloc() method now also checks for the existence of model parameters and allocates the appropriate model from the registry that is found in the XML file base on model type and parameter names. This allows now to have the same model type but different parameter names in the registry. The spectral model classes have been adapted to the new logic, and models have been added to the registry for all parameter combinations that may exist. The unit tests and documentation have been updated.
-
…eval_gradients() methods.
-
The value constructor now takes values for all 4 parameter members as arguments. The pivot and cut-off energy arguments have been declared as GEnergy. All parameter members now have set and retrieve methods. The pivot and cut-off method set and retrieve GEnergy objects. A precomputation cache has been added for the eval() and eval_gradients() methods that does accelerate computations in case that the energy and the index, cutoff and pivot energy parameters have not changed. This should improve the performance for binned analysis.
-
The eval_gradients() method is no longer const to reflect the fact that the method explicitly sets the gradient attribute of the model parameters.
-
The autoscale() method has been removed as it exists now in the GModelSpectral() base class. The parameter constructor now calls autoscale() to perform an autoscaling of all model parameters.
