                 Changes in version 2.17.2 (2026-08-20)                 

    o   Bug fixes in R code.

                 Changes in version 2.17.1 (2025-11-10)                 

    o   Refactored code for EM algorithm calculation.

    o   Added three different methods for EM acceleration calculation.

    o   Added squared update for EM iteration acceleration.

    o   Added aitken accelerated log likelihood estimation.

    o   Added new slot likelihood.tolerance.check to the class
	EM.control.

    o   Added new slot likelihood.estimation.rule to the class
	EM.control.

    o   Removed bug from EM maximization step for normal and lognormal
	mixtures.

    o   Improved C++ code for rough and enhanced Weibull and gamma
	parameter estimation.

    o   New C++ function BayesWeibullParameters().

                 Changes in version 2.16.0 (2024-07-10)                 

    o   Some minor changes and bug fixes in C++ and R codes.

    o   Improved C++ code for handling warnings and errors.

    o   Improved C++ functions RoughEstimationKNN(),
	RoughEstimationKDE() and RoughEstimationH().

    o   Additional argument Mode enters function REBMIX.

    o   Improved C++ functions GlobalModeKNN(), GlobalModeKDE() and
	GlobalModeH().

                 Changes in version 2.15.0 (2023-07-16)                 

    o   Method "labelmoments" is added.

    o   Method "mergelabels" is added.

    o   Adjacency matrix A slot is added to "RCLRMIX" class.

    o   Improved C++ functions CombineComponentsEntropy() and
	CombineComponentsDemp().

                 Changes in version 2.14.2 (2022-08-17)                 

    o   Valgrind problems have been resolved.

                 Changes in version 2.14.1 (2022-07-21)                 

    o   Class "Histogram" is added.

    o   Methods "chistogram", "fhistogram" and mapclusters are added.

    o   Argument y0 is removed from REBMIX, bins, optbins methods and
	from class "REBMIX".

    o   The C++ function REBMIXK() is added. It is called by the REBMIX
	method if the argument Dataset contains objects of type
	"Histogram".

    o   Numerous minor changes and debugging in C++ and R codes
	required due to the above changes.

    o   Methods "demix", "dfmix", "pemix", "pfmix" and all information
	criterion methods are modified for class "Histogram".

                 Changes in version 2.14.0 (2022-02-02)                 

    o   The package has been extended to three parametric family types.

    o   Improved C++ functions REBMIXKNN(), REBMIXKDE() and REBMIXH().

    o   Method plot is debugged and improved.

    o   Value "cdf" is added to what argument.

    o   Class "EMMIX.Theta" is added.

    o   Method EMMIX is added.

                 Changes in version 2.13.1 (2021-07-28)                 

    o   Reduced size of the vignette.

    o   Conditional jump or move depends on uninitialised value(s.

                 Changes in version 2.13.0 (2021-07-14)                 

    o   Method plot is debugged and improved.

    o   Improved C++ functions REBMIXKNN(), REBMIXKDE() and REBMIXH().

    o   EM algorithm is added to the C++ code for the improved
	estimation of all mixture model parameters.

    o   Updated examples.

                 Changes in version 2.12.0 (2020-06-08)                 

    o   Value "density" is replaced by "pdf" in what argument.

    o   Value "marginal" is replaced by "marginal pdf" in what
	argument.

    o   Value "distribution" is replaced by "marginal cdf" in what
	argument.

    o   Improved C++ functions regarding memory allocation and
	efficiency.

    o   Improved outlier detection.

    o   "Knuth equal" and "Knuth unequal" binning rules are added to
	the existing "Sturges", "Log10" and "RootN" rules.

    o   Method plot is improved.

    o   Method RCLRMIX is debugged and improved.

    o   Method REBMIX is debugged and improved.

    o   Method bins is added.

    o   Method optbins is added.

    o   Fixed bugs in C++ and R code.

                 Changes in version 2.11.0 (2019-12-11)                 

    o   Method split is improved.

    o   Improved C++ functions REBMIXKNN(), REBMIXKDE() and REBMIXH().

    o   Method RCLSMIX is improved.

    o   Method plot is improved.

    o   Continuous parametric family "Gumbel" is added.

    o   EM algorithm is added to the C++ code for the improved
	estimation of the Gaussian mixture model parameters.

    o   Class EM.Control is added for EM algorithm control parameters
	handling.

                 Changes in version 2.10.3 (2018-09-27)                 

    o   Method Zp is replaced by a.Zp accessor method.

    o   Method coef is replaced by a.theta1.all and a.theta2.all
	accessor methods.

    o   Accessor methods are added for all slots.

    o   Class "RNGMIX.Theta" is added.

    o   Method show is added for "RCLS.chunk" class.

    o   Minimum number of component cmin is added to REBMIX arguments.

    o   Minimum number of component cmin slot is added to "REBMIX"
	class.

    o   Improved C++ functions REBMIXKNN(), REBMIXKDE() and REBMIXH().

    o   Preprocessing "Parzen window" is renamed to "kernel density
	estimation".

    o   Rough parameter estimation for binomial and Poisson parametric
	families is improved.

    o   Latent class analysis is added.

                 Changes in version 2.10.2 (2018-05-22)                 

    o   Method split is improved.

    o   Fixed GCC 8.1 notes and warnings in C++ functions.

                 Changes in version 2.10.1 (2018-04-07)                 

    o   Fixed numerical problems related to high dimensional datasets.

    o   Fixed bug in REBMIX method.

    o   Improved C++ functions REBMIXKNN(), REBMIXPW() and REBMIXH().

    o   Improved C++ functions RoughEstimationKNN(),
	RoughEstimationPW() and RoughEstimationH().

                 Changes in version 2.10.0 (2018-03-03)                 

    o   Fixed bug in plot method.

    o   Additional argument rseed enters boot method.

    o   Maximum number of nonempty bins has been introduced in C++ to
	avoid histograms with excessive number of nonempty bins.

                 Changes in version 2.9.3 (2017-08-28)                  

    o   Acceleration rate has been improved. It is now progressively
	increasing.

    o   The Newton-Raphson root finding in C++ functions has been
	improved. It affects only "Weibull", "gamma" and "vonMises"
	parametric families.

                 Changes in version 2.9.2 (2017-06-09)                  

    o   Continuous circular parametric family "vonMises" is added.

    o   Fixed bug in C++ function REBMIXH.

    o   Improved C++ functions InformationCriterionKNN(),
	InformationCriterionPW() and InformationCriterionH().

    o   Improved C++ functions EnhancedEstimationKNN(),
	EnhancedEstimationPW() and EnhancedEstimationPW().

                 Changes in version 2.9.1 (2017-03-19)                  

    o   Properly registered native routines and disabled symbol search
	in C++ code.

                 Changes in version 2.9.0 (2017-01-26)                  

    o   Fixed bugs in RCLRMIX method.

                 Changes in version 2.8.4 (2016-12-08)                  

    o   Method Zp is added.

    o   Fixed bugs in plot method.

    o   Fixed bugs in split method.

    o   Fixed bugs in RCLRMIX method.

    o   Fixed bugs in RCLSMIX method.

    o   Improved C++ functions GlobalModeKNN(), GlobalModePW() and
	GlobalModeH().

    o   Improved C++ functions REBMIXKNN(), REBMIXPW() and REBMIXH().

                 Changes in version 2.8.3 (2016-10-05)                  

    o   Fixed convergence bug in C++ code.

    o   Bayes classification is improved by introducing outliers.

    o   Method plot is improved.

    o   Fixed bug in C++ functions RoughEstimationKNN(),
	RoughEstimationPW() and RoughEstimationH().

    o   Improved C++ functions GlobalModeKNN(), GlobalModePW() and
	GlobalModeH().

                 Changes in version 2.8.2 (2016-06-20)                  

    o   Fixed information criterion bug in C++ and R code.

    o   Fixed heap block freed with an inappropriate deallocation
	function bug in C++ code.

    o   Method plot is improved.

    o   Methods demix, dfmix, pemix and pfmix are improved.

    o   Methods split, chunk and BFSMIX are added.

    o   Renewed vignette.

                 Changes in version 2.8.1 (2016-03-27)                  

    o   Fixed isnan and isinf bug in C++ code.

    o   Fixed invalid object bug in vignette.

                 Changes in version 2.8.0 (2016-03-25)                  

    o   Object-oriented C++ code is improved considerably.

    o   R code is extended and rewritten in S4 class system.

    o   Multivariate normal parametric family with unrestricted
	variance-covariance matrices is enabled.

    o   Classification is improved.

    o   Clustering is added.

    o   Datasets iris is added.

    o   Renewed vignette.

    o   Fixed bug in C++ functions RoughLognormalParameters(),
	RoughWeibullParameters() and RoughGammaParameters().

                 Changes in version 2.7.2 (2015-08-13)                  

    o   Switch _TIME_LEFT_SWITCH is eliminated from C code.

    o   C code is converted into object-oriented C++ code.

    o   Fixed no visible global function definition bug in NAMESPACE
	file.

    o   Renewed vignette.

                 Changes in version 2.7.1 (2015-04-04)                  

    o   Demos rebmix.galaxy, rebmix.univariate.normal,
	rebmix.mixed.continuous.discrete and rebmix.multivariate.normal
	are changed.

                 Changes in version 2.7.0 (2015-02-10)                  

    o   Fixed title field in DESCRIPTION file.

    o   Fixed S3 generic/method consistency bug in NAMESPACE file.

    o   Fixed line endings in Makevars file.

    o   Fixed \VignetteKeyword entries bug in vignette.

    o   Method predict.list is converted to function RCLSMIX.

    o   Method boot is converted to function boot.REBMIX.

    o   Method plot is improved.

    o   Function kseq is improved.

    o   Function REBMIX is improved.

    o   Total of positive relative deviations D is eliminated from
	REBMIX arguments.

    o   Value "c" is eliminated from what argument.

    o   Value "K" is added to what argument.

    o   Fixed bug in C functions InformationCriterionKNN(),
	InformationCriterionPW() and InformationCriterionH().

    o   Fixed bug illegal characters found in URL.

    o   Methods AIC, AIC3, AIC4, AICc, CAIC, BIC, HQC, MDL2, MDL5, AWE,
	CLC, ICL, ICLBIC, PRD, SSE and PC are converted to functions
	with the same names.

    o   Renewed vignette.

                 Changes in version 2.6.2 (2014-09-05)                  

    o   Demos rebmix.galaxy, rebmix.univariate.normal,
	rebmix.mixed.continuous.discrete, rebmix.multivariate.normal,
	rebmix.weibullnormal and rebmix.truck are added.

    o   Datasets weibullnormal and truck are added.

    o   Method boot.REBMIX for bootstrap is improved.

    o   Function kseq for sequence of bins or nearest neighbours
	generation is added.

    o   Additional argument y0 enters function REBMIX.

    o   Improved treatment of accuracy Eps in C functions.

    o   Fixed bug in C functions EnhancedEstimationKNN(),
	EnhancedEstimationPW() and EnhancedEstimationH().

                 Changes in version 2.6.1 (2014-07-24)                  

    o   Renewed vignette.

    o   Renewed references in Reference manual.

    o   Method PD is added.

    o   Fixed bug in C functions REBMIXKNN(), REBMIXPW() and REBMIXH().

                 Changes in version 2.6.0 (2014-06-25)                  

    o   Switch _TIME_LEFT_SWITCH introduced in C code.

    o   Simplified treatment of loose restraints in C functions
	RoughEstimationKNN(), RoughEstimationPW() and
	RoughEstimationH().

    o   Fixed bug in C functions REBMIXKNN(), REBMIXPW() and REBMIXH().

    o   Minimum weight multiplier b is eliminated from REBMIX
	arguments.

    o   Fixed source and destination overlap in memcpy bug in C code.

                 Changes in version 2.5.1 (2013-10-12)                  

    o   Fixed checking top-level files.

    o   Fixed checking for old-style vignette sources.

    o   Packages mixtools and mclust excluded from vignette.

                 Changes in version 2.5.0 (2013-01-25)                  

    o   Rough parameter estimation is improved by introducing
	constrained entropy maximization.

    o   Fixed bug in C functions RoughEstimationKNN(),
	RoughEstimationPW() and RoughEstimationH().

    o   Fixed bug in boot.REBMIX method.

                 Changes in version 2.4.2 (2012-11-11)                  

    o   Fixed symbol eating bug for Mac OS X.

                 Changes in version 2.4.1 (2012-06-28)                  

    o   Continuous parametric family "gamma" is added.

                 Changes in version 2.4.0 (2012-04-23)                  

    o   Methods AIC, AIC3, AIC4, AICc and CAIC are added.

    o   Method AWE is added.

    o   Method BIC is added.

    o   Method boot.REBMIX for bootstrap is added.

    o   Method CLC is added.

    o   Function demix for empirical density calculation is added.

    o   Function dfmix for predictive density calculation is added.

    o   Method HQC is added.

    o   Method ICL is added.

    o   Method ICLBIC is added.

    o   Method logL is added.

    o   Methods MDL2 and MDL5 are added.

    o   Function pemix for empirical distribution function calculation
	is added.

    o   Function pfmix for predictive distribution function calculation
	is added.

    o   Function plot.REBMIX is extended. It enables besides "pdf"
	plots also "marginal pdf", "c", "IC", "logL", "D" and "marginal
	cdf" plots.

    o   Additional arguments what enters method plot.REBMIX.

    o   Method PRD is added.

    o   Methods print.boot.REBMIX and summary.boot.REBMIX are added.

    o   Method SSE is added.

    o   Dataset weibull is added.

                 Changes in version 2.3.0 (2012-01-25)                  

    o   Discrete parametric families "binomial", "Poisson" and "Dirac"
	are added.

    o   Information criteria "D" and "SSE" are added.

    o   Methods print.RNGMIX, print.REBMIX, coef.REBMIX and
	summary.REBMIX are added.

    o   Method predict.list for class membership prediction is added.

    o   The package does not write to stdout and read from stdout any
	more.

    o   The package is protected against wrong input arguments.

    o   Additional arguments Variables, theta1, theta2, ymin, ymax and
	b enter function REBMIX.

    o   Minimum radius RMIN is eliminated.

    o   Simplified treatment of global modes with equal probabilities
	in C functions GlobalModeKNN(), GlobalModePW() and
	GlobalModeH().

    o   Improved treatment of potentially singular solutions in C
	functions EnhancedEstimationKNN(), EnhancedEstimationPW() and
	EnhancedEstimationH().

    o   Improved convergence criteria in C functions REBMIXKNN(),
	REBMIXPW() and REBMIXH().

    o   Fixed bug in C function PreprocessingPW().

    o   Fixed bug in C functions RoughEstimationKNN(),
	RoughEstimationPW() and RoughEstimationH().

    o   Fixed bug in plot.REBMIX method.

    o   Dataset adult is added.
