Interact with models

You can interact with the model to tune a model from io data for dynamical system identification. In addition, a large number of parameters are available to select the model to be tuned and the optimization algorithm.

Tune a model

To tune and calibrate a model whether it is black box or grey box you can use the command :tune:

julia> model(:tune; 
        project_name = "project_1",
        model_name = "model_name",
        io = "io_name",
        computation_solver = "lls",
        model_architecture = "linear", 
    )

where project_name is the name of the project, model_name is the model name when saved on the database, io is the data used to tune the model, computation_solver is the solver used to tune the model, computation_maximum_time is the time allocated to tune the model and model_architecture is the model architecture.

The blackbox models which can be selected are visible in section blackbox model’s description :

Blackbox models description

During a model tuning, a hyperparameters optimization is performed with a meta_heuristic algorithm. More information is provided in the following section:

Hyperparameters optimization

Furthermore, in order to measure the identification performance of a model, a cost function is used, which is presented in the section:

Loss function

It is possible to add extra parameters to tune a model from io data, depending on blackbox model. All the parameters are listed on parameters section:

Parameters models

Finally, the available solvers to tune a model are depicted in solvers section:

Solvers

List models

You can list the models from a project:

julia> model(:ls, project_name = "project_1")

Remove a model

You can remove a model from a dedicated project:

julia> model(:rm, project_name = "project_1", model_name = "model_name")