src.baseline.bbo.random_search

Module Contents

Classes

Random_search

Introduction Random_search is an implementation of a basic random search optimization algorithm, inheriting from Basic_Optimizer. It generates random candidate solutions within the problem bounds and tracks the best solution found so far. The optimizer supports logging of progress and optional collection of full meta-data for analysis.

API

class src.baseline.bbo.random_search.Random_search(config)

Bases: src.environment.optimizer.basic_optimizer.Basic_Optimizer

Introduction

Random_search is an implementation of a basic random search optimization algorithm, inheriting from Basic_Optimizer. It generates random candidate solutions within the problem bounds and tracks the best solution found so far. The optimizer supports logging of progress and optional collection of full meta-data for analysis.

Initialization

Initializes the random search optimizer with the config object constructed in src/config.py.

Args:

• config (object):

  • The Attributes needed for the Random_search in config are the following:

    • maxFEs (int): Maximum number of function evaluations allowed.Defaullt value depends on the type of the problem.

    • n_logpoint (int): Number of log points for tracking progress. Default is 50.

    • log_interval (int): Interval at which logs are recorded.

    • full_meta_data (bool): Flag indicating whether to use full meta data.Default is None.

Attributes:

• __fes (int): Counter for the number of function evaluations performed.Default is 0.

• log_index (int or None): Index for logging, initialized as None.Default is None.

• cost (any): Placeholder for cost value, initialized as None.Default is None.

• __NP (int): Population size, set to 100.

__str__()

Returns a string representation of the Random Search optimizer.

Returns:

• str: The name of the optimizer, ‘Random_search’.

__reset(problem)

Introduction

Resets the internal state of the random search optimizer for a new optimization run on the given problem.

Args:

• problem: The optimization problem instance to initialize the population for.

Effects:

• Resets the function evaluation counter.

• Clears the cost history.

• Initializes a new random population.

• Appends the initial global best solution to the cost history.

• Sets the log index to 1.

__random_population(problem, init)

Introduction

Generates a random population of candidate solutions within the problem’s bounds, evaluates their costs, and updates the global best solution.

Args:

• problem (object): The optimization problem instance, expected to have attributes lb (lower bounds), ub (upper bounds), dim (dimension), optimum (optional optimum value), and an eval method for evaluating solutions.

• init (bool): Indicates whether this is the initial population generation. If True, initializes the global best; otherwise, updates it if a better solution is found.

Side Effects:

• Updates self.meta_Cost and self.meta_X if self.full_meta_data is True.

• Increments self.__fes by the population size (self.__NP).

• Updates self.gbest with the minimum cost found in the current population.

run_episode(problem)

Introduction

Executes a single optimization episode using random search on the provided problem instance. Tracks the best solution found, logs progress at specified intervals, and optionally collects metadata for analysis.

Args:

• problem: An object representing the optimization problem.

Returns:

• dict: A dictionary containing:

  • ‘cost’ (list): The logged best costs at each interval.

  • ‘fes’ (int): The total number of function evaluations performed.

  • ‘metadata’ (dict, optional): If self.full_meta_data is True, includes:

    • ‘X’ (list of np.ndarray): The population positions at each logging interval.

    • ‘Cost’ (list of float): The fitness values of the population at each logging interval.

Notes:

• The function ensures that the cost log is filled up to the required number of log points.

• If full_meta_data is True, additional metadata about the search process is included in the results.