Implementing Metaheuristics

MetaGen provides a flexible framework for developing and integrating new metaheuristic algorithms. Developers can implement metaheuristics in different ways depending on their needs and the level of integration required with the framework.

Developer Profiles

There are three main approaches to developing a metaheuristic in MetaGen, each with different levels of integration with the framework:

Standard Metaheuristic: The simplest approach, where the developer implements a metaheuristic using the standard Solution class without extending it. This case is illustrated in the MetaGen in Action section.
Flexible Metaheuristic: The developer ensures that the metaheuristic is compatible with possible future extensions of the Solution class by using a BaseConnector.
Extended Metaheuristic: The developer requires a custom Solution class for domain-specific operations and ensures that the metaheuristic remains compatible with standard and extended solutions.

Standard Metaheuristic

A Standard Metaheuristic is implemented using the built-in Solution class. This approach is ideal when the developer does not need custom data structures or variable types. The metaheuristic can be applied to any optimization problem without requiring modifications to the framework.

This case is covered in the MetaGen in Action section.

Flexible Metaheuristic

A Flexible Metaheuristic ensures compatibility with future extensions of the Solution class by obtaining the BaseConnector from the Domain and using it to instantiate solutions. This allows the metaheuristic to work seamlessly with both standard and extended Solution representations.

Example:

from metagen.framework import Domain

# Define a domain using a specific connector
connector = AnyConnector()
domain = Domain(connector)
domain.define_integer("max_depth", 2, 8)
domain.define_integer("n_estimators", 2, 16)

# Dynamically determine the correct solution type
solution_type: type[Solution] = domain.get_connector().get_type(domain.get_core())
potential: Solution = solution_type(domain, connector=domain.get_connector())

Key characteristics:

Uses BaseConnector from the Domain to instantiate solutions.
Ensures compatibility with both standard and extended Solution classes.
Future extensions of Solution can be used without modifying the metaheuristic.

Extended Metaheuristic

An Extended Metaheuristic is implemented when a developer needs a custom extension of the Solution class. This is necessary when the metaheuristic requires domain-specific data structures or specialized operations beyond what the standard Solution class offers.

Example (GA with a custom Solution class):

from metagen.framework import Domain
from metagen.metaheuristics.ga import GA
from metagen.metaheuristics.ga_tools import GAConnector, GASolution

# Define a domain using the GA-specific connector
connector = GAConnector()
domain = Domain(connector)
domain.define_integer("max_depth", 2, 8)
domain.define_integer("n_estimators", 2, 16)

# Dynamically determine the correct solution type
solution_type: type[Solution] = domain.get_connector().get_type(domain.get_core())
potential: Solution = solution_type(domain, connector=domain.get_connector())

Key characteristics:

Requires defining a custom Solution class (e.g., GASolution).
The metaheuristic must work with both standard and extended Solution classes.
Uses a custom BaseConnector (e.g., GAConnector) to map the new Solution class to the framework.

Choosing the Right Approach

Approach	Key Features	Best Suited For
Standard Metaheuristic	Uses standard Solution class Works with any Domain	General-purpose optimization without extensions
Flexible Metaheuristic	Uses BaseConnector for compatibility Can work with standard and extended Solution	Ensuring compatibility with extended solutions
Extended Metaheuristic	Requires a custom Solution class Uses a custom BaseConnector	Specialized metaheuristics with custom structures

MetaGen is designed primarily for developing Standard and Flexible metaheuristics. However, advanced users can extend the framework to create Extended metaheuristics when needed.

By selecting the appropriate approach, developers can ensure their metaheuristics are both efficient and compatible with future enhancements in MetaGen.