changelog-3-4-1

由java开发的软件包

2004-01-14 14:57  eibe

	* weka/core/Utils.java (1.38): logs2probs had unintended side
	effect.

2004-01-14 14:57  eibe

	* weka/clusterers/: ClusterEvaluation.java (1.22),
	DistributionClusterer.java (1.10), DistributionMetaClusterer.java
	(1.8), EM.java (1.23): Logification: things are numerically more
	stable now.

2004-01-14 12:07  eibe

	* weka/clusterers/DistributionClusterer.java (1.9): resurrected

2004-01-13 16:44  eibe

	* weka/filters/unsupervised/attribute/ClusterMembership.java (1.2):
	Some fixes.

2004-01-13 15:12  mhall

	* weka/gui/GenericObjectEditor.props (1.89): Added
	ClusterMembership filter

2004-01-13 15:11  mhall

	* weka/filters/unsupervised/attribute/ClusterMembership.java (1.1):
	Initial import

2004-01-12 17:28  mhall

	* weka/attributeSelection/OneRAttributeEval.java (1.14): Added
	options to control cross validation parameters and to allow the
	user to opt for using the training data to evaluate attribute
	goodness (rather than x-val). Can now also specify the minimum
	bucket size for OneR.

2004-01-10 20:39  eibe

	* weka/classifiers/meta/MultiClassClassifier.java (1.35): Pairwise
	classification works now if two classes are empty.

2004-01-09 14:00  mhall

	* weka/classifiers/functions/neural/NeuralNetwork.java (1.12):
	Fixed so that option click acts as right button on Macs

2004-01-09 12:09  eibe

	* weka/classifiers/functions/Logistic.java (1.31): Finished model
	uses a lot less memory now.

2004-01-09 11:17  mhall

	* weka/gui/treevisualizer/TreeVisualizer.java (1.7): Fixed so that
	option click acts as right button on Macs

2004-01-08 16:49  mhall

	* weka/clusterers/DistributionMetaClusterer.java (1.7): Now fits
	normal and discrete estimators to clusters produced by the wrapped
	clusterer

2004-01-08 16:47  mhall

	* weka/clusterers/DistributionClusterer.java (1.8): Removed from
	repository

2004-01-08 12:37  mhall

	* weka/gui/LogPanel.java (1.14): Fixed so that right click (using
	option/alt key) on Macs with one mouse button now works

2004-01-06 10:17  eibe

	* weka/classifiers/meta/MultiClassClassifier.java (1.34): Changed
	random code generation dependend on input data (i.e. seed is now
	based on an instance chosen from the input data).

2004-01-05 09:17  eibe

	* weka/: classifiers/meta/Decorate.java (1.1),
	classifiers/misc/FLR.java (1.1), gui/GenericObjectEditor.props
	(1.88): Added to classifiers that have been contributed:
	Decorate.java and FLR.java

2003-12-19 10:48  eibe

	* weka/filters/unsupervised/instance/RemoveMisclassified.java
	(1.2): Changed RemoveMisclassified to just remove instances from
	the first batch. Now it can be used together with the
	FilteredClassifier.

2003-12-18 11:28  eibe

	* weka/: core/Utils.java (1.37),
	experiment/RandomSplitResultProducer.java (1.16): Added method for
	probabilistic rounding in Utils.java and changed the
	RandomSplitResultsProducer to use this method. Previously there was
	the potential for slight bias in the class distribution due to
	rounding in the stratification phase.

2003-12-17 14:15  akibriya

	* weka/classifiers/bayes/ComplementNaiveBayes.java (1.2): Fixed a
	bug so it can be used without setting any options.

2003-12-16 17:26  eibe

	* weka/gui/: GUIChooser.java (1.12), ResultHistoryPanel.java
	(1.18), explorer/AssociationsPanel.java (1.17),
	explorer/AttributeSelectionPanel.java (1.30),
	explorer/ClassifierPanel.java (1.70), explorer/ClustererPanel.java
	(1.42), explorer/Explorer.java (1.28),
	explorer/PreprocessPanel.java (1.37), visualize/AttributePanel.java
	(1.9), visualize/ClassPanel.java (1.12), visualize/LegendPanel.java
	(1.4), visualize/MatrixPanel.java (1.7), visualize/Plot2D.java
	(1.20), visualize/VisualizePanel.java (1.20): Weka Explorer is now
	just Explorer.

2003-12-16 13:42  eibe

	* weka/classifiers/functions/supportVector/SMO.java (1.4): SMO
	should be quite a bit faster now: the data wasn't randomized after
	it was split into two-class problems (even if the original problem
	only had two classes).

2003-12-15 09:33  mhall

	* weka/attributeSelection/: BestFirst.java (1.23),
	CfsSubsetEval.java (1.19): Modified to improve memory usage on data
	sets with many features

2003-12-12 12:42  eibe

	* weka/gui/visualize/Plot2D.java (1.19): Changed access for some
	variables from private to protected.

2003-12-05 17:41  eibe

	* weka/core/Instances.java (1.49): Fixed a bug that I just
	introduced with my new code.

2003-12-05 16:39  eibe

	* weka/: attributeSelection/AttributeSelection.java (1.31),
	attributeSelection/OneRAttributeEval.java (1.13),
	attributeSelection/RaceSearch.java (1.12),
	attributeSelection/WrapperSubsetEval.java (1.21),
	classifiers/Evaluation.java (1.49),
	classifiers/evaluation/EvaluationUtils.java (1.9),
	classifiers/functions/supportVector/SMO.java (1.3),
	classifiers/meta/CVParameterSelection.java (1.23),
	classifiers/meta/Grading.java (1.3),
	classifiers/meta/LogitBoost.java (1.28),
	classifiers/meta/MultiScheme.java (1.14),
	classifiers/meta/OrdinalClassClassifier.java (1.10),
	classifiers/meta/RandomCommittee.java (1.3),
	classifiers/meta/Stacking.java (1.18),
	classifiers/meta/StackingC.java (1.5),
	classifiers/meta/ThresholdSelector.java (1.29),
	classifiers/rules/ConjunctiveRule.java (1.9),
	classifiers/rules/JRip.java (1.12), classifiers/rules/Ridor.java
	(1.9), classifiers/rules/part/MakeDecList.java (1.12),
	classifiers/rules/part/PART.java (1.20),
	classifiers/trees/REPTree.java (1.15),
	classifiers/trees/RandomTree.java (1.6),
	classifiers/trees/j48/J48.java (1.28),
	classifiers/trees/j48/PruneableClassifierTree.java (1.8),
	clusterers/ClusterEvaluation.java (1.21), clusterers/EM.java
	(1.22), core/Instances.java (1.48),
	experiment/CrossValidationResultProducer.java (1.13),
	gui/beans/CrossValidationFoldMaker.java (1.4),
	gui/explorer/AttributeSelectionPanel.java (1.29),
	gui/explorer/ClassifierPanel.java (1.69): Cross-validation now
	randomizes the data within each training fold before it is passed
	to the learning scheme. This is important for schemes that are
	sensitive to the order in the training data (in particular, those
	ones that create a random number generator based on a selected
	instance in the training data). Previously we have assumed that the
	Classifier itself will randomize the data if necessary. (The same
	change applies to various meta-classifiers, etc, as well.)

2003-12-05 13:49  eibe

	* weka/filters/supervised/attribute/ClassOrder.java (1.3): Re-wrote
	large chunks of ClassOrder.

2003-12-01 17:10  akibriya

	* weka/classifiers/bayes/NaiveBayesMultinomial.java (1.6):
	Commented out the factorial terms used in calculating
	prob(word|class), as it doesn't make any difference in classifier's
	results and is unnecessary.

2003-11-28 12:12  eibe

	* weka/classifiers/BVDecomposeSegCVSub.java (1.1): Initial import.

2003-11-27 17:31  akibriya

	* weka/gui/GenericObjectEditor.props (1.87): Added RandomProjection
	filter.

2003-11-27 17:30  akibriya

	* weka/filters/unsupervised/attribute/RandomProjection.java (1.1):
	Initial import.

2003-11-27 13:43  akibriya

	* weka/core/Stopwords.java (1.1): Initial import.

2003-11-27 13:39  akibriya

	* weka/filters/unsupervised/attribute/StringToWordVector.java
	(1.6): Added a range of options to complement the new
	ComplementNaiveBayes classifier.

2003-11-27 13:35  akibriya

	* weka/gui/GenericObjectEditor.props (1.86): Added
	ComplementNaiveBayes classifier.

2003-11-27 13:34  akibriya

	* weka/classifiers/bayes/ComplementNaiveBayes.java (1.1): Initial
	import.

2003-11-27 12:14  mhall

	* weka/classifiers/trees/m5/: Rule.java (1.7), RuleNode.java (1.6):
	Fixed a bug relating to minNumInstances

2003-11-26 09:54  eibe

	* weka/gui/experiment/ResultsPanel.java (1.25): Fixed bug

2003-11-24 17:38  eibe

	* weka/gui/experiment/ResultsPanel.java (1.24): Introduced a split
	panel in the results panel.

2003-11-24 09:01  eibe

	* weka/gui/explorer/ClassifierPanel.java (1.68): Relative errors
	for cross-validation were different from command-line estimates
	because calculation of the mean was not done based on each training
	fold.

2003-11-24 08:57  eibe

	* weka/classifiers/trees/j48/C45PruneableClassifierTree.java
	(1.11): Fixed bug in J48. Empty leaves didn't get flagged as
	non-empty if they became populated after subtree raising had
	occurred. This could affect probability estimates for instances
	with missing values (in cases where leaves were empty before
	pruning, subtree raising actually occurred, and the missing
	attribute was tested immediately above the previously empty leaf).

2003-11-20 16:38  eibe

	* weka/classifiers/trees/j48/: BinC45Split.java (1.8),
	C45Split.java (1.8): Made a change that affects probability
	estimates for empty leaves.

2003-11-20 11:15  eibe

	* weka/classifiers/meta/RegressionByDiscretization.java (1.23):
	Fixed bug: didn't use to work for datasets with missing class
	values

2003-11-19 14:03  eibe

	* weka/: classifiers/meta/RegressionByDiscretization.java (1.22),
	filters/unsupervised/attribute/Discretize.java (1.6): Another small
	fix.

2003-11-19 11:11  eibe

	* weka/filters/: supervised/attribute/Discretize.java (1.3),
	unsupervised/attribute/Discretize.java (1.5),
	unsupervised/attribute/PotentialClassIgnorer.java (1.2): Fixed a
	bug. Also: split points are now put half-way between values in
	supervised.attribute.Discretize.

2003-11-18 17:45  eibe

	* weka/: classifiers/functions/supportVector/SMOreg.java (1.4),
	classifiers/meta/RegressionByDiscretization.java (1.21),
	filters/Filter.java (1.23),
	filters/unsupervised/attribute/Discretize.java (1.4),
	filters/unsupervised/attribute/Normalize.java (1.3),
	filters/unsupervised/attribute/NumericToBinary.java (1.2),
	filters/unsupervised/attribute/PotentialClassIgnorer.java (1.1),
	filters/unsupervised/attribute/ReplaceMissingValues.java (1.3),
	filters/unsupervised/attribute/Standardize.java (1.3): Made it
	possible for some unsupervised attribute filters to ignore class
	(or not).

2003-11-17 14:08  eibe

	* weka/: core/Instances.java (1.47),
	filters/supervised/attribute/Discretize.java (1.2),
	filters/unsupervised/instance/RemoveWithValues.java (1.4): Removed
	numeric padding in tests.

2003-11-14 11:10  eibe

	* weka/classifiers/meta/AdditiveRegression.java (1.14): Yesterday's
	fix wasn't quite the right thing. Fixed it properly (and made it
	faster, too :-).

2003-11-13 17:31  eibe

	* weka/classifiers/meta/AdditiveRegression.java (1.13): Fixed bug
	that resulted in reduced performance when shrinkage parameter was
	set to value different from 1.

2003-11-13 12:03  eibe

	* weka/: classifiers/meta/AttributeSelectedClassifier.java (1.14),
	classifiers/meta/CVParameterSelection.java (1.22),
	classifiers/meta/CostSensitiveClassifier.java (1.17),
	classifiers/meta/FilteredClassifier.java (1.19),
	classifiers/meta/ThresholdSelector.java (1.28), core/Drawable.java
	(1.7): Made a few more meta classifiers implement Drawable.

2003-11-12 14:06  mhall

	* weka/build.xml (1.11): Updated build file

2003-11-12 13:27  eibe

	* weka/: Makefile (1.40), associations/Makefile (1.4),
	attributeSelection/Makefile (1.16), classifiers/Makefile (1.30),
	classifiers/bayes/Makefile (1.8), classifiers/evaluation/Makefile
	(1.6), classifiers/functions/Makefile (1.8),
	classifiers/functions/neural/Makefile (1.2),
	classifiers/functions/pace/Makefile (1.3),
	classifiers/functions/supportVector/Makefile (1.2),
	classifiers/lazy/Makefile (1.6), classifiers/lazy/kstar/Makefile
	(1.4), classifiers/meta/Makefile (1.8), classifiers/misc/Makefile
	(1.2), classifiers/rules/Makefile (1.6),
	classifiers/rules/part/Makefile (1.2), classifiers/trees/Makefile
	(1.6), classifiers/trees/adtree/Makefile (1.2),
	classifiers/trees/j48/Makefile (1.4),
	classifiers/trees/lmt/Makefile (1.2), classifiers/trees/m5/Makefile
	(1.4), clusterers/Makefile (1.6), core/Makefile (1.14),
	core/converters/Makefile (1.2), datagenerators/Makefile (1.2),
	estimators/Makefile (1.3), experiment/Makefile (1.14),
	filters/Makefile (1.27), filters/supervised/Makefile (1.2),
	filters/supervised/attribute/Makefile (1.2),
	filters/supervised/instance/Makefile (1.2),
	filters/unsupervised/Makefile (1.2),
	filters/unsupervised/attribute/Makefile (1.6),
	filters/unsupervised/instance/Makefile (1.2), gui/Makefile (1.22),
	gui/beans/Makefile (1.4), gui/boundaryvisualizer/Makefile (1.3),
	gui/experiment/Makefile (1.6), gui/explorer/Makefile (1.7),
	gui/graphvisualizer/Makefile (1.2), gui/streams/Makefile (1.3),
	gui/treevisualizer/Makefile (1.2), gui/visualize/Makefile (1.6):
	Removed all Makefiles because we use ant to compile weka. There is
	no point in maintaining them.

2003-11-12 11:39  eibe

	* weka/classifiers/meta/Bagging.java (1.25): Changed bagging so
	that weights are properly taken into account even if out-of-bag
	error is calculated.

2003-11-12 11:07  eibe

	* weka/classifiers/functions/LeastMedSq.java (1.8): Behaviour of
	RemoveRange has changed.

2003-11-12 11:00  mhall

	* weka/gui/beans/FilterBeanInfo.java (1.3): Javadoc fix

2003-11-12 10:13  eibe

	* weka/classifiers/: meta/CVParameterSelection.java (1.21),
	trees/RandomTree.java (1.5): Two tiny changes.