Package org.uma.evolver.metaoptimizationproblem

Class MetaOptimizationProblem<S extends org.uma.jmetal.solution.Solution<?>>

java.lang.Object

org.uma.jmetal.problem.doubleproblem.impl.AbstractDoubleProblem

org.uma.evolver.metaoptimizationproblem.MetaOptimizationProblem<S>

Type Parameters:

S - The type of solutions used by the base algorithm being optimized

All Implemented Interfaces:

Serializable, org.uma.jmetal.problem.doubleproblem.DoubleProblem, org.uma.jmetal.problem.Problem<org.uma.jmetal.solution.doublesolution.DoubleSolution>

public class MetaOptimizationProblem<S extends org.uma.jmetal.solution.Solution<?>>
extends org.uma.jmetal.problem.doubleproblem.impl.AbstractDoubleProblem

A meta-optimization problem that optimizes the parameters of an optimization algorithm by evaluating its performance across multiple problem instances using quality indicators.

This class implements a meta-optimization approach where the parameters of a base algorithm are automatically tuned by evaluating its performance on multiple problem instances using various quality indicators. The optimization objective is to find parameter settings that work well across all problem instances in the training set.

Key features:

• Supports optimization of any configurable algorithm that implements BaseLevelAlgorithm
• Allows evaluation on multiple problem instances with different characteristics
• Uses quality indicators to assess algorithm performance
• Supports multiple independent runs to account for stochasticity
• Provides flexible evaluation budget control through EvaluationBudgetStrategy

Typical usage involves:

1. Define the base algorithm with its parameter space
2. Select a set of training problems
3. Configure quality indicators for evaluation
4. Set up the evaluation budget strategy
5. Use an optimization algorithm to solve the meta-optimization problem

See Also:

• EvaluationBudgetStrategy
• BaseLevelAlgorithm
• QualityIndicator
• Serialized Form

Field Summary

Fields inherited from class org.uma.jmetal.problem.doubleproblem.impl.AbstractDoubleProblem

bounds, name, numberOfConstraints, numberOfObjectives

Constructor Summary

Constructors

Constructor	Description
MetaOptimizationProblem(BaseLevelAlgorithm<S> baseAlgorithm, List<org.uma.jmetal.problem.Problem<S>> problems, List<String> referenceFrontFileNames, List<org.uma.jmetal.qualityindicator.QualityIndicator> indicators, EvaluationBudgetStrategy evaluationBudgetStrategy, int numberOfIndependentRuns)	Constructs a new meta-optimization problem instance.

Method Summary

Modifier and Type	Method	Description
org.uma.jmetal.solution.doublesolution.DoubleSolution	evaluate(org.uma.jmetal.solution.doublesolution.DoubleSolution solution)	Evaluates a solution by running the base algorithm with the specified parameter settings and computing the quality indicators.
EvaluationBudgetStrategy	evaluationBudgetStrategy()	Returns the evaluation strategy being used.
String	name()	Returns the name of the problem.
int	numberOfConstraints()	Returns the number of constraints in the problem.
int	numberOfObjectives()	Returns the number of objectives in the problem.
int	numberOfVariables()	Returns the number of variables (parameters) in the problem.
List<Parameter<?>>	parameters()	Returns the list of parameters being optimized.

Methods inherited from class org.uma.jmetal.problem.doubleproblem.impl.AbstractDoubleProblem

createSolution, name, numberOfConstraints, numberOfObjectives, variableBounds, variableBounds

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Constructor Details

MetaOptimizationProblem

public MetaOptimizationProblem(BaseLevelAlgorithm<S> baseAlgorithm,
 List<org.uma.jmetal.problem.Problem<S>> problems,
 List<String> referenceFrontFileNames,
 List<org.uma.jmetal.qualityindicator.QualityIndicator> indicators,
 EvaluationBudgetStrategy evaluationBudgetStrategy,
 int numberOfIndependentRuns)

Constructs a new meta-optimization problem instance.

Parameters:

baseAlgorithm - the base algorithm whose parameters will be optimized (must not be null)

problems - the list of problems to evaluate the algorithm on (must not be null or empty)

referenceFrontFileNames - list of file paths containing reference fronts for each problem (must match the size of problems list)

indicators - list of quality indicators to evaluate solutions (must not be null or empty)

evaluationBudgetStrategy - strategy for determining the evaluation budget for each problem (must not be null and must be compatible with the number of problems)

numberOfIndependentRuns - number of independent runs to perform for each evaluation (must be positive)

Throws:

NullPointerException - if any parameter is null

IllegalArgumentException - if the sizes of problems and referenceFrontFileNames don't match, if the evaluation budget strategy is not compatible with the number of problems, or if numberOfIndependentRuns is not positive

See Also:

• EvaluationBudgetStrategy.validate(int)

Method Details

numberOfVariables

public int numberOfVariables()

Returns the number of variables (parameters) in the problem.

Specified by:

numberOfVariables in interface org.uma.jmetal.problem.Problem<S extends org.uma.jmetal.solution.Solution<?>>

Overrides:

numberOfVariables in class org.uma.jmetal.problem.doubleproblem.impl.AbstractDoubleProblem

Returns:

the number of parameters being optimized

numberOfObjectives

public int numberOfObjectives()

Returns the number of objectives in the problem.

Specified by:

numberOfObjectives in interface org.uma.jmetal.problem.Problem<S extends org.uma.jmetal.solution.Solution<?>>

Overrides:

numberOfObjectives in class org.uma.jmetal.problem.doubleproblem.impl.AbstractDoubleProblem

Returns:

the number of quality indicators being used

numberOfConstraints

public int numberOfConstraints()

Returns the number of constraints in the problem. This implementation returns 0 as the problem is unconstrained.

Specified by:

numberOfConstraints in interface org.uma.jmetal.problem.Problem<S extends org.uma.jmetal.solution.Solution<?>>

Overrides:

numberOfConstraints in class org.uma.jmetal.problem.doubleproblem.impl.AbstractDoubleProblem

Returns:

0 (no constraints)

name

public String name()

Returns the name of the problem.

Specified by:

name in interface org.uma.jmetal.problem.Problem<S extends org.uma.jmetal.solution.Solution<?>>

Overrides:

name in class org.uma.jmetal.problem.doubleproblem.impl.AbstractDoubleProblem

Returns:

the name of the problem as a string

evaluationBudgetStrategy

public EvaluationBudgetStrategy evaluationBudgetStrategy()

Returns the evaluation strategy being used.

Returns:

the evaluation strategy

parameters

public List<Parameter<?>> parameters()

Returns the list of parameters being optimized.

Returns:

an unmodifiable list of parameters

evaluate

public org.uma.jmetal.solution.doublesolution.DoubleSolution evaluate(org.uma.jmetal.solution.doublesolution.DoubleSolution solution)

Evaluates a solution by running the base algorithm with the specified parameter settings and computing the quality indicators.

Parameters:

solution - the solution containing the parameter values to evaluate

Returns:

the evaluated solution with objective values set

Throws:

NullPointerException - if solution is null

IllegalArgumentException - if the solution's variables don't match the expected number of parameters