dinjo.optimizer module

class dinjo.optimizer.Optimizer(model: dinjo.model.ModelIVP, reference_state_variable: dinjo.model.StateVariable, reference_values: List[float], reference_t_values: List[float], integration_method: str = 'RK45')[source]

Bases: object

Optimizes the initial value problem’s parameters, as defined in the class ModelIVP.

model

the initial value problem to be optimized.

Type

ModelIVP

reference_state_variable :class:`StateVariable`

the state variable to be fitted to the solution of the IVP.

reference_values

the ‘experimental data’ of the reference state variable to be used as the fitting variable.

Type

list[float]

reference_t_values

the corresponding times at which the reference_values are given.

Type

list[float]

integration_method

must be one of the methods accepted by scipy.integrate.solve_ivp.

Type

str

__init__(model: dinjo.model.ModelIVP, reference_state_variable: dinjo.model.StateVariable, reference_values: List[float], reference_t_values: List[float], integration_method: str = 'RK45') → None[source]

Initialize self. See help(type(self)) for accurate signature.

cost_function(parameters: List[float], cost_method: str = 'root_mean_square')[source]

Function to be minimized by the optimizer. Initially this will be the root mean square of the difference between the observations and the numerical solution.

Parameters

  • parameters (list[float]) – parameters of the model to be minimized. The order of the parameters must be the same as they appear in self.model.parameters.

  • cost_method (str) – Must be one of ['root_mean_square',].

optimize(cost_method: str = 'root_mean_square', algorithm: str = 'differential_evolution', algorithm_kwargs: Dict[str, Any] = {}) → scipy.optimize.optimize.OptimizeResult[source]

Global minimization of cost function.

Parameters

  • cost_function_method (str) – Must be one of the permitted values for cost_method parameter in Optimize.cost_function().

  • algorithm (str) – scipy.optimize algorithm used for the optimization. Must be one of 'differential_evolution', 'shgo', 'dual_annealing'.

  • algorithm_kwargs (dict[str, any]) – parameters passed to the optimization algorithm. They are different depending on the algorithm.

Returns

object containing the minimization result as returned by the appropiate scipy algorithm (e.g. if the chosen al).

Return type

minimization

property reference_t_values