Recent changes to bugs

LBestTopology inherits GBestTopology iterator()

Thu, 30 Apr 2009 06:28:26 -0000

The LBestTopology.iterator() method is not implemented, meaning the functionality is inherited from GBestTopology which returns a GBestNeighbourhoodIterator instead of an LBestNeighbourhoodIterator.

GA maximises FunctionMinimisationProblem

Tue, 21 Apr 2009 07:24:06 -0000

The ec.EC algorithm configured with GeneticAlgorithmIterationStrategy is maximising problem that should be minimised.

FitnessCalculator only wants Entity

Fri, 13 Feb 2009 06:44:40 -0000

Someone commented out the FitnessCalculator in the clustering.KMeans class, because FitnessCalculator now requires an Entity and KMeans only has a Vector. I don't recall when this change in FitnessCalculator was made. The Git logs don't contain the change.

KMeans is not a population based algorithm and therefore an Entity would not make sense. Is there another type of FitnessCalculator that can handle Vectors?

Because the fitness is not calculated, the clusters never get created and the following exception occurs:

./simulator.sh xml/kmeans.xml
Starting simulation 1 of 1.
log4j:WARN No appenders could be found for logger (net.sourceforge.cilib.util.ClusteringUtils).
log4j:WARN Please initialize the log4j system properly.
clusters: null
Exception in thread "Thread-1" java.lang.NullPointerException
at net.sourceforge.cilib.clustering.kmeans.KMeans.algorithmIteration(KMeans.java:113)
at net.sourceforge.cilib.algorithm.Algorithm.performIteration(Algorithm.java:141)
at net.sourceforge.cilib.algorithm.Algorithm.run(Algorithm.java:183)
at java.lang.Thread.run(Thread.java:619)

Bug at RandomAlgorithmIterator

Olusegun — Wed, 08 Oct 2008 23:11:52 -0000

The class doesn't work... Likely because the list of random numbers is not being correctly instantiated.

A file that works is attached...

Subtle bug in Entity.equals() when convergence occurs

Gary Pampara — Fri, 03 Oct 2008 16:16:17 -0000

When the scenario of all entities converging onto a single point in the search space, the .equals() and .hashCode() methods for Entity objects fail. All objects are suddenly the same. This causes a problem within the parentSelection of the DE, resulting in an infinite loop.

Redundancy in BoundaryConstraint classes...

Olusegun — Fri, 12 Sep 2008 14:47:55 -0000

There are two different implementations of the boundary constraint handling mechanisms for pso.

One of them defines an enforce(...) interface and can be found in the pso.iterationstrategies package while the other defines a constrainLower(...) and constrainUpper(...) interafce and can be found in pso.positionupdatestrategies.boundaryconstraintstrategies...

The former is built into the iteration strategy and provides a more "elegant" solution (this is my opinion). I, thus, suggest that the implementation of the latter be merged into it...

Regards,

Arnith

TournamentSelectionStrategy returns worst entity...

Olusegun — Fri, 29 Aug 2008 15:59:53 -0000

The strategy should return the winner of the tournament, right? As far as I can tell, this is the individual with the best fitness...

The comparator in the select method sorts according to descending fitness (i.e. from worst to best) and returns the worst (individual at index 0)...

Suggested fix: comparator should be changed from DescendingFitness to AscendingFitness...

I doubt I'm missing anything here but pardon me if I am...

Regards...

___

Killearnith

FDRVelocityUpdate breaks on Ackley

Olusegun — Thu, 28 Aug 2008 19:24:34 -0000

The velocity update in question either causes stagnation on Ackley or very slow minimization.

Original paper proposing the strategy should be pulled up and the velocity strategy should be re-implemented...

Killearnith

SelectiveContributionUpdateStrategy incorrectly implemented

Olusegun — Tue, 19 Aug 2008 09:17:26 -0000

This class is supposed to update part of a cooperative entity (context vector) with the solution represented by participant algorithm entity only if the updated context is better than the original.

Currently, the comparison is between the participant entity and the context instead of between the original context and the updated context. This needs to be corrected and tested.

Objects don't implement equals & hashCode methods correctly

Gary Pampara — Sun, 22 Jun 2008 21:12:21 -0000

Within a number of objects, the equals() and hashCode() methods are not correct.

Many of the equals() methods do not test for reference equality or nulls and the hashCode() should be written to match the equals().