canari.optimizer

This module automates the search for optimal hyperparameters of Model and SKF instances by leveraging the external libraries Ray Tune and Optuna.

class canari.optimizer.Optimizer(model: Callable, param: dict, model_input: dict | None = None, num_optimization_trial: int | None = 50, grid_search: bool | None = False, mode: str | None = 'min', algorithm: str | None = 'TPE', back_end: str | None = 'ray', num_startup_trials: int | None = 20)

Bases: object

Optimize hyperparameters for Model and SKF using the Ray Tune and Optuna external libraries. Optimization is based on the metric saved in metric_optim for Model or metric_optim for SKF.

Parameters:

• model (Callable) – Function that returns a model instance given a model configuration.

• param (Dict[str, list]) – Parameter search space: two-value lists [min, max] for defining the bounds of the optimization. Users can also use Ray Tune search space object such as: tune.randint(12, 53), tune.uniform(0.1, 0.4), tune.loguniform(1e-1, 4e-1).

• model_input (Dict) – Any other inputs for the model that is different from the model’s parameters.

• num_optimization_trial (int, optional) – Number of random search trials (ignored for grid-search). Defaults to 50.

• grid_search (bool, optional) – If True, perform grid search. Defaults to False.

• algorithm (str, optional) – Search algorithm: ‘TPE’ (OptunaSearch) or ‘random’ (random sampling). Defaults to ‘TPE’.

• mode (str, optional) – Direction for optimization stopping: “min” or “max”. Defaults to “min”.

• back_end (str, optional) – “ray”. Using the external library Ray for optimization.

• num_startup_trials (int, optional) – Number of start up trial when using TPE sampling. Defaults to 20.

model_optim

The best model instance initialized with optimal parameters after running optimize().

param_optim

The best hyperparameter configuration found during optimization.

Type:

Dict

get_best_model()

Retrieve the optimized model instance after running optimization.

Returns:

Model instance initialized with the best hyperparameter values.

get_best_param() → Dict

Retrieve the optimized parameters after running optimization.

Returns:

Best hyperparameter values.

Return type:

dict

objective(config: Dict) → Dict

Returns a metric that is used for optimization

Returns:

Metric used for optimization.

Return type:

dict

optimize()

Run optimziation