Base Module¶
This module provides a suite of reusable components for building and implementing evolutionary algorithms. It is designed to facilitate the development of novel optimization strategies by providing a foundation of common classes and interfaces.
Algorithm¶
- class pyevopt.base.Algorithm¶
The Algorithm class is an abstract base class that defines the basic interface for all evolutionary algorithms.
This class is designed to help developers create customized algorithms logic for their evolutionary algorithms. By implementing this class, users can define new algorithms to use in the optimization process.
- next(population, objective)¶
This method applies the optimization strategy defined by this algorithm to the given population, using the provided objective function as the fitness metric.
Parameters¶
- population: Population
Represent the current generation of candidate solutions.
- objective: Callable
A callable function that evaluates the quality of each individual in the population.
Returns¶
- pyevopt.base.Population:
The new population of candidate solutions after a single iteration of the genetic algorithm.
- dict:
A dictionary containing information about the current iteration.
Raises¶
- NotImplementedError
If the method is not implemented in a subclass.
Notes¶
This method is intended to be called repeatedly to iterate through the optimization process.
The specific behavior of this method depends on the concrete subclass and its implementation of the evolutionary algorithm.
It is assumed that the objective function is deterministic, meaning it always returns the same output for a given input.
This method is abstract and should be implemented in subclasses. The exact behavior depends on the specific algorithm used by the subclass.
EarlyStopCriteria¶
- class pyevopt.base.EarlyStopCriteria(threshold, init_value)¶
This class provides a way to control the optimization process by determining when to stop prematurely based on certain conditions.
This class is designed to help developers create customized early stopping logic for their evolutionary algorithms. By implementing this class, users can define specific criteria for terminating the optimization process.
Attributes¶
- threshold: float
Boundary value that, when reached or exceeded, signals to stop the optimization process prematurely.
- init_value: float
Starting point of the optimization process. It represents the first value that will be evaluated by the algorithm.
- value: float
Represents the latest evaluation result produced by the optimization process. This attribute serves as a reference point for monitoring progress and detecting convergence or exceeding the maximum threshold.
- check() bool¶
Check the condition to decide if a new iteration of the optimization process should be executed.
Return¶
- bool:
True if stop criteria is meet, False otherwise.
Raises¶
- NotImplementedError
If the method is not implemented in a subclass.
Notes¶
This method is abstract and should be implemented in subclasses. The exact behavior depends on the specific algorithm used by the subclass.
- init()¶
Initializes the EarlyStopCriteria object by resetting the current value (value) to the initial value (init_value). This sets the starting point for evaluating the optimization process.
- next(**kwargs) None¶
Update the current value for the next iteration of the optimization process.
Parameters¶
- **kwargs: dict, optional
evaluation: Number of objective function evaluation that are executed by the algorithm.
Raises¶
- NotImplementedError
If the method is not implemented in a subclass.
Notes¶
This method is abstract and should be implemented in subclasses. The exact behavior depends on the specific algorithm used by the subclass.
Sampler¶
- class pyevopt.base.Sampler¶
This class provides a way to generate new populations.
This class is designed to help developers create customized samplers logic to generate new populations. By implementing this class, users can define new ways to generate it.
- generate(size: int, dimension: int, domain)¶
Generate a population of individuals.
Generates a population of size individuals, each individual being an array of length dimension, with values chosen from the specified domain.
Parameters¶
- size: int
The number of individuals in the population.
- dimension: int
The size of each individual.
- domain
A function that maps a value to a random sample from this domain.
Returns¶
- np.ndarray
An array of shape (size, dimension) containing the generated population.
Raises¶
- NotImplementedError
If the method is not implemented in a subclass.
Notes¶
This method is abstract and should be implemented in subclasses. The exact behavior depends on the specific algorithm used by the subclass.
Population¶
- class pyevopt.base.Population(population, domain)¶
Class representing a population of individuals.
This class holds the information about a population of individuals. It stores the actual population, its fitness values, number of individuals, dimension of each individual and domain from which the individuals were generated.
Attributes¶
- population: np.ndarray
The array containing the population.
- fitness: np.ndarray
The array containing the fitness values for each individual.
- n: int
The number of individuals in the population.
- dimension: int
The size of each individual.
- domain:
A function that maps a value to a random sample from this domain.
- evaluate(objective)¶
Evaluate each individual in the population using the provided objective function.
This method calculates the fitness value for each individual in the population by applying the provided objective function to each individual. The results are stored in the fitness attribute of this object.
Parameters¶
- objective: callable
A function that takes an individual and returns its fitness value.
Notes¶
After calling this method, you can access the fitness values for each individual using the fitness attribute.