src.environment.problem.SOO.HPO_B.hpo_b

Module Contents

Classes

HPOB_Problem

Introduction

HPO-B is an autoML hyper-parameter optimization benchmark which includes a wide range of hyperparameter optimization tasks for 16 different model types (e.g., SVM, XGBoost, etc.), resulting in a total of 935 problem instances. The dimension of these problem instances range from 2 to 16. We also note that HPO-B represents problems with ill-conditioned landscape such as huge flattern.

API

class src.environment.problem.SOO.HPO_B.hpo_b.HPOB_Problem(bst_surrogate, dim, y_min, y_max, lb, ub, normalized=False, name=None)[source]

Bases: src.environment.problem.basic_problem.Basic_Problem

Introduction

HPO-B is an autoML hyper-parameter optimization benchmark which includes a wide range of hyperparameter optimization tasks for 16 different model types (e.g., SVM, XGBoost, etc.), resulting in a total of 935 problem instances. The dimension of these problem instances range from 2 to 16. We also note that HPO-B represents problems with ill-conditioned landscape such as huge flattern.

Original paper

“Hpo-b: A large-scale reproducible benchmark for black-box hpo based on openml.” arXiv preprint arXiv:2106.06257 (2021).

Official Implementation

HPO-B

License

None

Initialization

Initializes the class with the provided surrogate model, dimensionality, bounds, and other configuration parameters.

Args:

  • bst_surrogate: The surrogate model to be used for optimization.

  • dim (int): The dimensionality of the problem.

  • y_min (float): The minimum value of the objective function.

  • y_max (float): The maximum value of the objective function.

  • lb (array-like): The lower bounds for each dimension.

  • ub (array-like): The upper bounds for each dimension.

  • normalized (bool, optional): Whether to normalize the input data. Defaults to False.

  • name (str, optional): The name of the problem instance. Defaults to None.

Built-in Attributes:

  • bst_surrogate: Stores the surrogate model.

  • y_min: Stores the minimum objective value.

  • y_max: Stores the maximum objective value.

  • dim: Stores the problem dimensionality.

  • gbest: Stores the global best value found so far.Defaults to 1e+10.

  • normalized: Indicates if normalization is applied.

  • collect_gbest: List to collect global best values during optimization.

  • lb: Stores the lower bounds.

  • ub: Stores the upper bounds.

  • optimum: Stores the optimum solution (if found).

  • name: Stores the name of the problem instance.

func(position)[source]

Introduction

Evaluates the surrogate model prediction for a given position, normalizes the result, and updates the global best collection.

Args:

  • position (np.ndarray): The input position vector to be evaluated, expected to have shape compatible with (self.dim,).

Returns:

  • float: The negative normalized prediction from the surrogate model for the given position.

Side Effects:

  • Appends the current global best (self.gbest) to the self.collect_gbest list.

Notes:

  • Uses an XGBoost surrogate model (self.bst_surrogate) for prediction.

  • Normalization is performed using the self.normalize method.

normalize(y)[source]

Introduction

Normalizes the input array y to the range [0, 1] based on the object’s normalization settings.

Args:

  • y (np.ndarray or float): The input value(s) to be normalized.

Returns:

  • np.ndarray or float: The normalized value(s) if normalization is enabled; otherwise, returns the input as is.

Notes:

  • If self.normalized is True and self.y_min is None, normalization is performed using the min and max of y.

  • If self.normalized is True and self.y_min is set, normalization uses self.y_min and self.y_max and clips the result to [0, 1].

  • If self.normalized is False, the input y is returned unchanged.

__str__()[source]

Introduction

Returns the string representation of the object, typically its name.

Returns:

  • str: The name attribute of the object.