| Type |
Changes |
By |
|---|
 |
Corrected many issues regarding generics. The code should be cleaner
and have less type warnings. |
gpampara |
|
Converted the function shifting into a function decorator and added the
ability to rotate functions. |
gpampara |
 |
Implemented a BoundedPositionUpdateStrategy to adjust Particle behaviour at search
space boundaries. Six different boundary constraint strategies have been added:
BouncingBoundaryConstraintStrategy, ClampingBoundaryConstraintStrategy, DeflectionBoundaryConstraintStrategy, NearestBoundaryConstraintStrategy, PeriodicBoundaryConstraintStrategy and RandomBoundaryConstraintStrategy |
heatzync |
|
Fixed some potential bugs in ControlParameterUpdateStrategy code. |
gpampara |
|
Added the new structure for the diversity measuement. |
gpampara |
|
Corrected the manner in which the structure of the PSO iteration strategies are defined (ASynchrounsousIterationStrategy and SynchronousIterationStrategy). The code should be more clear. |
gpampara |
|
Improved the implementation and usage of the problem.dataset.AssociatedPairDataSetBuilder
and functions.continuous.ClusteringFitnessFunction hierarchies. |
heatzync |
|
functions.continuous.ScatterSeperationRatio.java is a generalisation for validity indices that
make use of the intra-cluster (within-cluster) scatter and inter-cluster (between-cluster)
seperation values. These include the Davies-Bouldin index as well as the Dunn indices. |
heatzync |
|
Two of the Generalised Dunn Indices have been implemented: functions.continuous.DunnIndex33.java
and functions.continuous.DunnIndex53.java. functions.continuous.GeneralisedDunnIndex.java was
added to help with the implementation. |
heatzync |
|
functions.continuous.MaulikBandyopadhyayIndex.java: a new Clustering Validity Index
functions.continuous.VeenmanReindersBackerIndex.java: a new Clustering Validity Index |
heatzync |
![]() |
functions.continuous.HalkidiVazirgiannisIndex.java was previously called functions.continuous.ScatterDensityIndex.java. |
heatzync |
![]() |
problem.dataset.CachingDataSetBuilder.java was previously called problem.dataset.CachedDistanceDataSetBuilder.java. |
heatzync |
|
measurement.single.NumberOfClustersFormed.java is a new measurement that measures the number
of clusters that are formed by the best solution of the current iteration. |
heatzync |
|
Updated and re-implemented unit tests. |
heatzync |
 |
Removed functions.continuous.QuantisationErrorTest.java, because its functionality is tested
inside functions.continuous.ClusteringFitnessFunctionTest.java. |
heatzync |
|
measurement.single.GenericFunctionMeasurement.java: A measurement that takes any function
(you have to use the .setFunction() method) and returns the value (or fitness) of that
function. This class was created so that it is possible to make use of any of the existing
functions in the "functions" package without the need to actually optimise that function,
i.e. without using it as a fitness function. |
heatzync |
|
pso.iterationstrategies.PerElementReinitialisation.java: A boundary constraint that only
reinitialises those components of the Vector (inside the Entity) that have move outside of
the bounds, as opposed to reinitialising the entire Entity as in the case of the
ReinitialisationBoundary class. |
heatzync |
|
Refactored a few classes. Corrected some typos. Added some more fitness functions for intra-cluster measurements. |
heatzync |
|
pso.iterationstrategies.BouncingPositionUpdateStrategy.java: Instead of making use of
boundary constraints that reinitialise entire Particles (or components thereof), this class
is a proactive approach to prevent the Particles from moving outside of the domain. Before
moving a Particle to it's new location, this position update strategy first checks to see
whether the Particle will be outside of the domain. If not, the Particle is moved. If one of
the components of a Particle will be outside of the domain, it is placed very close to the
boundary of the domain (but still inside) and the Particle's velocity for that component is
inverted (multiplied by -1), effectively making the Particle bounce of the sides of the
domain. |
heatzync |
|
Added a reference to the Problem from within the DataSetBuilder. |
heatzync |
|
Some classes used Double.MIN_VALUE incorrectly. Double.MIN_VALUE is the smallest double
(that can be represented in Java) that is greater than 0.0. The smallest double is actually
-Double.MAX_VALUE. |
heatzync |
|
Added the revised Entity heirarchy. This effectively allows for dynamic
properties to be associated with all Entity objects. The main driver
for the change was the need for extra information with regards to
coevolution changes that will come into CILib shortly. Entity
implementations such as Particle, Individual and the like now share
a datastructure to store their properties within. |
gpampara |
|
Added the new classes to the Type system. This change effectively
enables the Entity hierarchy change by moving the container classes
such as Tree into the Type system so that they can be used within
Entities.
A further modification was the notion that all Trees, Sets and Vectors
are essentially special cases of Grpahs. This now means that all
container types extend from Graph interface. The old container
code will remain until all the functionality has been ported to the
new packages. |
gpampara |
|
Refactored some code and made some methods generic. Cleaned up some
comments and mainly made sure the tree is currently in a sane state. |
gpampara |
|
Renamed a few methods to make the intention of the methods more clear. |
gpampara |
|
Changed the Type system hierarchy by pushing the isInsideBounds() method
up into the Type class. This change simplified the DomainParser code. |
gpampara |
|
Removed the need for a user defined exception class, as the
description is the main need and not the new class. |
gpampara |
