Package org.moeaframework.problem.MaF
Class MaF4
java.lang.Object
org.moeaframework.problem.AbstractProblem
org.moeaframework.problem.DTLZ.DTLZ
org.moeaframework.problem.MaF.MaF4
- All Implemented Interfaces:
AutoCloseable
,Problem
,AnalyticalProblem
The MaF4 test problem, which is an inverted and badly scaled version of the DTLZ3 problem. This problem exhibits
the following properties:
- Concave Pareto front
- Multimodal
- Badly scaled decision variables
- No single optimal solution in any subset of the objectives
-
Field Summary
Fields inherited from class org.moeaframework.problem.AbstractProblem
numberOfConstraints, numberOfObjectives, numberOfVariables
-
Constructor Summary
ConstructorDescriptionMaF4
(int numberOfObjectives) Constructs an MaF4 test problem with the specified number of objectives. -
Method Summary
Methods inherited from class org.moeaframework.problem.DTLZ.DTLZ
g1, g2, generateAt, getName, newSolution
Methods inherited from class org.moeaframework.problem.AbstractProblem
close, getNumberOfConstraints, getNumberOfObjectives, getNumberOfVariables
Methods inherited from class java.lang.Object
clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait
Methods inherited from interface org.moeaframework.core.Problem
close, getName, getNumberOfConstraints, getNumberOfObjectives, getNumberOfVariables, isType, newSolution
-
Constructor Details
-
MaF4
public MaF4(int numberOfObjectives) Constructs an MaF4 test problem with the specified number of objectives.- Parameters:
numberOfObjectives
- the number of objectives for this problem
-
-
Method Details
-
evaluate
Description copied from interface:Problem
Evaluates the solution, updating the solution's objectives in place. Algorithms must explicitly call this method when appropriate to evaluate new solutions or reevaluate modified solutions. -
generate
Description copied from interface:AnalyticalProblem
Returns a randomly-generated solution using the analytical solution to this problem. The exact behavior of this method depends on the implementation, but in general (1) the solutions should be non-dominated and (2) spread uniformly across the Pareto front.It is not always possible to guarantee these conditions. For example, a discontinuous / disconnected Pareto surface could generate dominated solutions, and a biased problem could result in non-uniform distributions. Therefore, we recommend callers filter solutions through a
NondominatedPopulation
, in particular one that maintains a spread of solutions.Furthermore, some implementations may not provide the corresponding decision variables for the solution. These implementations should indicate this by returning a solution with
0
decision variables.- Specified by:
generate
in interfaceAnalyticalProblem
- Returns:
- a randomly-generated Pareto optimal solution to this problem
-