Base Optimizer

BaseOptimizer

The BaseOptimizer class is the foundation for all optimization algorithms in Hyperactive v5.

Class Signature

class BaseOptimizer(BaseObject):
    def get_search_config(self) -> dict: ...
    def get_experiment(self) -> "BaseExperiment": ...
    def solve(self) -> dict: ...
    def _solve(self, experiment, *args, **kwargs) -> dict: ...  # to override

Key Methods

solve()

Run the optimization to maximize the experiment’s score (higher-is-better convention). Sets best_params_ and may set additional attributes depending on the optimizer.

Returns: - dict: Best parameters found

get_search_config()

Returns the optimizer’s configuration, excluding the experiment object (useful to introspect/store settings).

Returns: - dict: Optimizer-specific parameters

get_experiment()

Returns the associated experiment. If a plain callable was provided, it is wrapped as a FunctionExperiment automatically.

Returns: - BaseExperiment: Experiment to optimize

Properties

• best_params_ (dict): Best parameters found (set by solve())
• best_score_ (float, optional): Signed score per higher-is-better convention (set by some optimizers, e.g., GridSearchSk, RandomSearchSk)
• best_index_ (int, optional): Index of the best candidate in the evaluated sequence (set by some optimizers)

Minimal Example

from hyperactive.experiment.integrations import SklearnCvExperiment
from hyperactive.opt import GridSearchSk

exp = SklearnCvExperiment(...)
opt = GridSearchSk(param_grid={"C": [0.1, 1, 10]}, experiment=exp)
best = opt.solve()

Tags System

BaseOptimizer uses tags to describe algorithm properties:

• info:local_vs_global: local | mixed | global
• info:explore_vs_exploit: explore | exploit | mixed
• info:compute: low | middle | high