Openturns

=== Library ===

==== Major changes ====

==== New classes ====

==== API changes ====

=== Documentation ===

=== Python module ===

=== Miscellaneous ===

=== Library ===

==== Major changes ====
* New Gaussian process regression classes
* New method to compute the conditional expectation of a functional chaos expansion
* New class to obtain a conditional distribution wrt some components

==== New classes ====
* GaussianProcessFitter, GaussianProcessFitterResult (openturns.experimental)
* GaussianProcessRegression, GaussianProcessRegressionResult (openturns.experimental)
* GaussianProcessConditionalCovariance (openturns.experimental)
* PointConditionalDistribution (openturns.experimental)
* DistributionValidation (openturns.testing)

==== API changes ====
* SmolyakExperiment left the experimental module
* GeneralizedExtremeValueValidation left the experimental module
* TruncatedOverMesh left the experimental module
* StudentCopula left the experimental module
* StandardSpaceCrossEntropyImportanceSampling, PhysicalSpaceCrossEntropyImportanceSampling left the experimental module
* Removed deprecated method Study.printLabels
* Removed deprecated class TimerCallback
* Removed deprecated MetaModelValidation(inputSample, outputSample, metaModel) ctor
* Removed deprecated MetaModelValidation::computePredictivityFactor method
* Removed deprecated Solver.set/getMaximumFunctionEvaluation
* Removed deprecated Solver.getUsedFunctionEvaluation
* Removed deprecated Sample/CorrelationAnalysis.computePearsonCorrelation
* Removed deprecated Cobyla/TNC.setIgnoreFailure method
* Removed deprecated OptimizationAlgorithm.setMaximumEvaluationNumber method
* Removed deprecated OptimizationResult.getEvaluationNumber
* Deprecated BayesDistribution in favor of JointByConditioningDistribution
* Deprecated ConditionalDistribution in favor of DeconditionedDistribution
* Deprecated NegativeBinomial in favor of Polya
* Deprecated NegativeBinomialFactory in favor of PolyaFactory
* Deprecated OptimizationAlgorithm.Build(str) in favor of GetByName
* Removed EfficientGlobalOptimization.get/setImprovementFactor improvement factor accessors
* SobolSimulationAlgorithm.setExperimentSize must be used to set the DOE size instead of setBlockSize
* Deprecated SobolSimulationAlgorithm.setBatchSize, use setBlockSize instead
* Swapped InverseGamma shape/scale parameters: InverseGamma(k, lambda)

=== Documentation ===
* Copy button for code blocks
* Added new use case: Fission gas release and example of use

=== Python module ===
* Numpy 2 compatibility
* Workaround for matplotlib 3.6.x issues

=== Miscellaneous ===
* The Distribution PDF, logPDF and CDF can be exposed as Function objects using the new getPDF, getLogPDF and getCDF methods, previously they could only be computed on a Point or Sample or else drawn.
* Added a new constructor Student(nu, mu, Covariance)
* Added DistributionValidation class to automate distribution tests
* Cross-cuts of multivariate Functions can be drawn with the drawCrossCut method
* VisualTest::DrawInsideOutside shows which points in a Sample belong to a given Domain and which do not.
* Viridis replaces HSV as default colormap
* Add a new method VisualTest::DrawPairsXY to plot X vs Y samples
* New OrthogonalFunctionFactory.getMarginal method
* New EnumerateFunction.getMarginal method
* New FunctionalChaosResult.getConditionalExpectation

=== Bug fixes (total 39) ===
* 1077 (Docstrings suggest kwargs are available)
* 1179 (The SobolSimulationAlgorithm class has doc issues.)
* 1199 (LinearModelTest_LinearModelDurbinWatson does not take the firstSample into account in the actual Durbin-Watson test)
* 1203 (SpaceFillingMinDist documentation)
* 1241 (Wishart.getCovariance segfault)
* 1302 (EGO Example should show convergence)
* 1336 (The "Estimate a probability with FORM" example is unclear)
* 1417 (The example of the BarPlot graph is too complicated)
* 1631 (Missing image)
* 1739 (There is no example for a ProcessSample created from a Sample)
* 1748 (There is no worked example of the GaussKronrod algorithm)
* 1749 (There is no example or feature to plot a (X,Y) sample.)
* 1767 (The FORM explained example can be improved)
* 2166 (Curious indents on doc)
* 2318 (The target-HSIC example provides an incorrect algorithm parameterization)
* 2508 (Cobyla claim to support bounds, but does not)
* 2549 (SpecFunc.MaxScalar is undocumented)
* 2552 (ComputeQuantile for discrete variables)
* 2560 (CalibrationResult has no isBayesian() method)
* 2565 (give star discrepancy formula)
* 2629 (There are duplicate and inconsistent mathematical notations)
* 2674 (Add a "copybutton" feature to any code box in the doc)
* 2685 (Minor spelling errors in OT 1.23)
* 2689 (t_Bonmin_std.py fails on arm64, ppc64, ppc64el)
* 2690 (allow finish optimization wo feasible points)
* 2694 (InverseGamma and Gamma)
* 2706 (Doc "copy" feature is not robust to ellipses)
* 2707 (errors in Wilks doc)
* 2708 (NegativeBinomialDistribution should be renamed PolyaDistribution)
* 2710 (SimplicialCubature to use IntegrationAlgorithm interface)
* 2731 (The API help doc of CleaningStrategy is wrong)
* 2732 (MixtureClassifier.grade has a wrong error message)
* 2755 (The improvement factor of the EfficientGlobalOptimization is useless)
* 2761 (The name of the attributes is not consistent in the use cases)
* 2769 (Error when using a ConditionedGaussianProcess with output dimension > 1)
* 2773 (NAIS and StandardSpaceCrossEntropyImportanceSampling cannot deal with intersections of threshold events)
* 2780 (Student from (nu, mu, Covariance) ?)
* 2785 (The draw() method requires the same number of points in the X and Y directions)
* 2786 (The colors of some examples are broken)

=== Library ===

==== Major changes ====
* New GPD estimation services: MLE, profiled likelihood, time-varying, return level, covariates, clustering
* New GEV covariates estimation method, profiled likelihood by blocks
* Rank-based Sobol indices estimation
* Vector=>Field chaos-based metamodel and sensitivity algorithm

==== New classes ====
* FunctionalChaosValidation (openturns.experimental)
* SmoothedUniformFactory (openturns.experimental)
* GeneralizedParetoValidation (openturns.experimental)
* SamplePartition (openturns.experimental)
* PointToFieldFunctionalChaosAlgorithm (openturns.experimental)
* CubaIntegration (openturns.experimental)
* ExperimentIntegration (openturns.experimental)
* RankSobolSensitivityAlgorithm (openturns.experimental)
* UniformOrderStatistics (openturns.experimental)

==== API changes ====
* Deprecated MetaModelValidation(inputSample, outputSample, metaModel) is deprecated in favor of MetaModelValidation(outputSample, metamodelPredictions)
* Deprecated MetaModelValidation::computePredictivityFactor method, use computeR2Score
* Deprecated MetaModelValidation::getInputSample method
* Removed deprecated Pagmo.setGenerationNumber
* Removed deprecated NLopt.SetSeed
* Removed deprecated IterativeThresholdExceedance(dimension, threshold) ctor
* Removed deprecated Linear|QuadraticLeastSquares(Sample, Function) constructor
* Removed deprecated SubsetSampling.setKeepEventSample, getEventInputSample, getEventOutputSample
* Removed deprecated SubsetSampling.setISubset, setBetaMin
* Removed deprecated LHS
* Removed deprecated OptimizationAlgorithm.set/getVerbose
* Removed deprecated DickeyFullerTest.setVerbose/getVerbose
* Removed deprecated MetropolisHasting.setVerbose/getVerbose
* Removed deprecated BasisSequenceFactory.setVerbose/getVerbose
* Removed deprecated SimulationAlgorithm.setVerbose/getVerbose
* Removed deprecated WhittleFactory.setVerbose/getVerbose
* Removed deprecated CLassifier.setVerbose/getVerbose
* Removed deprecated ARMALLHF.setVerbose/getVerbose
* Removed deprecated CleaningStrategy.setVerbose/getVerbose
* Removed deprecated ApproximationAlgorithm.setVerbose/getVerbose
* Removed deprecated EllipticalDistribution.setCorrelation/getCorrelation
* Removed deprecated EllipticalDistribution.setMean
* Removed deprecated Distribution.computeDensityGenerator
* Removed deprecated Point.clean
* Removed deprecated (Inverse)Gamma.setKLambda
* Removed deprecated Os::GetEndOfLine
* Deprecated Sample|CorrelationAnalysis::computePearsonCorrelation in favor of computeLinearCorrelation
* Deprecated Cobyla::setIgnoreFailure, TNC::setIgnoreFailure in favor of setCheckStatus
* Deprecated OptimizationAlgorithm.set/getMaximumEvaluationNumber in favor of set/getMaximumCallsNumber
* Deprecated OptimizationResult.set/getEvaluationNumber in favor of set/getCallsNumber
* Deprecated Solver.set/getMaximumFunctionEvaluation in favor of set/getMaximumCallsNumber
* Deprecated Solver.getUsedFunctionEvaluation in favor of getCallsNumber
* Deprecated TimerCallback
* Deprecated ComposedDistribution in favor of JointDistribution
* Deprecated ComposedCopula in favor of BlockIndependentCopula
* Added Polygon::FillBetween static method to fill the surface between two curves
* Deprecated Study.printLabels (see getLabels)
* Removed optional parameters from Contour constructors, use set methods or ResourceMap keys to set them

=== Documentation ===
* Examples show how to visualize matrices
* Enforce check of internal links with sphinx nitpicky option

=== Python module ===
* Graphs apply default colors to Drawables with no explicit color
* Contour plots can now be filled and come with colorbars
* Binary wheels are now compatible with uv package manager

=== Miscellaneous ===
* CovarianceModel nuggetFactor can be optimized by KrigingAlgorithm
* UniformOverMesh left the experimental module
* IntegrationExpansion/LeastSquaresExpansion left the experimental module
* Allow to set optimization/simulation algorithm maximum run time duration
* New OptimizationAlgorithm API to retrieve return code and error message
* Improved TruncatedDistribution to use CDF inversion instead of rejection method to generate n-d samples
* Faster marginal distribution PDF computation by integration on marginalized components
* Extendend the JointDistribution to use any distribution defined on the unit cube that is not a copula

=== Bug fixes (total 53) ===
* 1252 (In NumericalMathFunction class, getCallsNumber and getEvaluationCallsNumber return the same information API)
* 1430 (The MetaModelValidation class has no graphics)
* 2067 (Can't compute of a marginal of a BayesDistribution->it takes ages!)
* 2218 (Split keepIntact methods)
* 2332 (The doc of Sobol' indices has issues)
* 2359 (The Sample help page does not show how to set a column)
* 2361 (Inconsistency in the documentation of KrigingAlgorithm)
* 2364 (KrigingAlgorithm: examples set bounds before disabling optimization)
* 2366 (The maths notations in the help pages are inconsistent)
* 2427 (Rename ComposedDistribution to JointDistribution ?)
* 2446 (The polynomial_sparse_least_squares theory help page can be improved)
* 2460 (Cobyla returns an internal exception when maximum number of evaluations is reached)
* 2474 (Coupling tools replace method is not robust to inputs larger than 10)
* 2477 (Deprecate Sample.computeLinearCorrelation ?)
* 2479 (Number of calls to objective function due to gradient approximation not always counted in OptimizationResult.getEvaluationNumber())
* 2489 (Inconsistent Evaluation number in drawOptimalValueHistory())
* 2500 (MarshallOlkinCopula missing computePDF ?)
* 2507 (The computeComplementaryCDF and computeSurvivalFunction methods are unclear)
* 2508 (Cobyla claim to support bounds, but does not)
* 2511 (Drop SpecFunc::IsNaN/IsInf)
* 2513 (Binomial quantile computation fails on extreme example)
* 2517 (Cannot build OpenTURNS with doc if optional dependencies bison/flex are unavailable)
* 2524 (Bayes distribution order)
* 2525 (Geometric range starts at 1)
* 2526 (BestModelChiSquared does not handle exceptions)
* 2527 (Strange probabilistic model for the fire satellite use-case)
* 2531 (StandardDistributionPolynomialFactory produces NaN and Infs)
* 2535 (undocumented SaltelliSensitivityAlgorithm model argument)
* 2541 (The computeQuantile() method of a Mixture can fail)
* 2544 (Example of multi output Kriging on the fire satellite model: paramers are not optimized)
* 2545 (Formatting Issues on LatentVariableModel)
* 2548 (Distribution::computeQuantile(p) can be computed for p<0 or p>1)
* 2552 (ComputeQuantile for discrete variables)
* 2557 (Shipping openturns and poissoninv GPL license)
* 2558 (Typos in Copulas theory documentation)
* 2563 (SimulatedAnnealingLHS.generate does not terminate for LHS of size 1)
* 2567 (The PythonRandomVector.getSample() method returns a 0-dimension sample)
* 2570 (MarginalDistribution test is disabled)
* 2571 (Drawables order is not preserved by the viewer)
* 2593 (Formatting of input and output arguments in the API help page)
* 2596 (The legends of the drawPDF() graph have a wrong order)
* 2602 (The pretty-print of a ParametricFunction does not show the name of the parameters)
* 2604 (Slower MonteCarlo simulations for versions > 1.19)
* 2621 (TruncatedDistribution n-d CDF inversion)
* 2624 (HSICEstimatorImplementation : cannot save with pickle)
* 2627 (Weird window for NormalGamma plot in API page)
* 2628 (Multi Start have incoherent behavior if the max eval has been reached)
* 2642 (The description of a KernelSmoothing fitted distribution can be lost sometimes)
* 2647 (Contour: need to detail the norms)
* 2653 (The Faure sequence is wrong)
* 2655 (HistogramFactory struggles with samples with various scales)
* 2658 (FunctionImplementation::draw forces the location of the color bar in 2D)
* 2673 (The invariant distribution of a DiscreteMarkovChain is wrong)

=== Library ===

==== Major changes ====

==== New classes ====
* BoundaryMesher (openturns.experimental)
* LatentVariableModel (openturns.experimental)
* StudentCopula (openturns.experimental)
* StudentCopulaFactory (openturns.experimental)
* TruncatedOverMesh (openturns.experimental)
* SimplicialCubature (openturns.experimental)

==== API changes ====
* Removed deprecated (Non)LinearLeastSquaresCalibration::getCandidate method
* Removed deprecated (Non)GaussianLinearCalibration::getCandidate method
* Removed deprecated DomainIntersection|Union|DisjunctiveUnion(Domain, Domain) ctors
* Removed deprecated EllipticalDistribution::getInverseCorrelation method
* Removed deprecated Basis::getDimension method
* Removed deprecated HSICStat ctors relying on weight matrix
* Deprecated Pagmo.setGenerationNumber for setMaximumIterationNumber
* Deprecated NLopt.SetSeed for setSeed
* Deprecated IterativeThresholdExceedance(dimension, threshold) ctor
* Deprecated Linear|QuadraticLeastSquares(Sample, Function) constructor
* Deprecated SubsetSampling.setKeepEventSample, getEventInputSample, getEventOutputSample
* Deprecated SubsetSampling.setISubset, setBetaMin
* QuantileMatchingFactory probabilities argument is no longer optional
* Add a new moment order argument to MethodOfMomentsFactory
* Removed Drawable.getPointCode
* Deprecated LHS in favor of ProbabilitySimulationAlgorithm+LHSExperiment
* Deprecated OptimizationAlgorithm.setVerbose/getVerbose
* Deprecated DickeyFullerTest.setVerbose/getVerbose
* Deprecated MetropolisHastings.setVerbose/getVerbose
* Deprecated BasisSequenceFactory.setVerbose/getVerbose
* Deprecated SimulationAlgorithm.setVerbose/getVerbose
* Deprecated WhittleFactory.setVerbose
* Deprecated Classifier.setVerbose/getVerbose
* Deprecated ARMAlikelihoodFactory.setVerbose/getVerbose
* Deprecated CleaningStrategy.setVerbose/getVerbose
* Deprecated ApproximationAlgorithm.setVerbose/getVerbose
* Deprecated EllipticalDistribution.setCorrelation/getCorrelation
* Deprecated EllipticalDistribution.setMean
* Student.setMu/getMu now operate on Point
* Deprecated Distribution.computeDensityGenerator methods
* Deprecated Point.clean
* Deprecated (Inverse)Gamma.setKLambda
* Deprecated Os::GetEndOfLine
* Moved PosteriorDistribution to experimental

=== Documentation ===
* Improved coverage of class/methods
* New two-degree-of-freedom oscillator use-case

=== Python module ===
* New Graph.setLegendCorner method to set legend outside of Graph
* Allowing use of matplotlib markers and legend location strings in Graphs

=== Miscellaneous ===
* Improved pretty-printing of chaos functions, distributions
* Added IterativeThresholdExceedance::getRatio
* Added FunctionalChaosResult::drawSelectionHistory to plot LARS coefs paths
* New API in SubsetSampling to get samples at each iteration
* Added SubsetSampling method to set the initial experiment
* Allowing use of non-independent copulas in LHSExperiment
* Improved system events to allow more kinds of events like DomainEvent
* Add CMake presets file

=== Bug fixes ===
* 1338 (The example in UserDefined does not show how to create a uniform discrete distribution)
* 1670 (IntervalMesher segfault when diamond==true)
* 1896 (Operator == inconsistencies)
* 1979 (The doc of KernelSmoothing has issues)
* 2195 (SubsetSampling: keep the failed and secure points at each step)
* 2216 (Characteristic function of the Triangular and Trapezoidal distributions)
* 2220 (Negative value given by the computeComplementaryCDF method)
* 2235 (Documentation is missing for some methods of class Matrix)
* 2339 (FunctionalChaosSobolIndices has doc issues)
* 2347 (Build error of /usr/include/tbb/machine/gcc_generic.h:39:20: error: operator '||' has no left operand in s390x on Fedora)
* 2351 (Import error in the draw method example)
* 2354 (IterativeThresholdExceedance)
* 2371 (t-Copula implementation)
* 2403 (KrigingAlgorithm fails if basis is empty)
* 2406 (Student does not give access to all its parameters)
* 2412 (The class QuadraticLeastSquares returns a wrong quadratic term)
* 2420 (Deprecate Os::GetEndOfLine())
* 2423 (mutable OptimizationAlgorithm)
* 2429 (Elliptical distributions)
* 2431 (LHSExperiment.generate() can fail)
* 2434 (WeibullMaxMuSigma: default values are not good)
* 2437 (Binomial computeSurvivalFunction error)
* 2439 (Use more computeScalarQuantile)
* 2442 (EGO does not handle maximization problems)
* 2443 (C++ Test SmolyakExperiment_std fails on arm64, ppc64el, s390x)
* 2452 (The LHSExperiment does not manage a non-independent copula)
* 2456 (configure fails to find bonmin)
* 2459 (Problems while configuring OpenTURNS for Visual Studio 2019)
* 2463 (Normal.computeConditionalPDF() is wrong when the components are independent)
* 2466 (PlackettCopula covariance matrix)
* 2481 (Improve SpectralGaussianProcess sampling speed)
* 2486 (DrawParallelCoordinates is not correct)
* 2487 (Cannot save large datasets using XMLH5StorageManager)
* 2503 (Using KernelSmoothing can make Jupyter Notebook to hang)

=== Library ===

==== Major changes ====
* New GEV estimation services: MLE, profiled likelihood, r-maxima, time-varying, return level

==== New classes ====
* SmolyakExperiment (openturns.experimental)
* Physical|StandardSpaceCrossEntropyImportanceSampling, CrossEntropyResult (openturns.experimental)
* LeastSquaresExpansion, IntegrationExpansion (openturns.experimental)
* UniformOverMesh (openturns.experimental)
* GeneralizedExtremeValueValidation (openturns.experimental)
* Coles (openturns.usecases.coles)
* Linthurst (openturns.usescases.linthurst)

==== API changes ====
* Deprecated LinearLeastSquaresCalibration::getCandidate, use getStartingPoint
* Deprecated NonLinearLeastSquaresCalibration::getCandidate, use getStartingPoint
* Deprecated GaussianLinearCalibration::getCandidate, use getParameterMean
* Deprecated NonGaussianLinearCalibration::getCandidate, use getParameterMean
* Removed SequentialStrategy
* Removed TensorApproximationAlgorithm, TensorApproximationResult, CanonicalTensorEvaluation, CanonicalTensorGradient
* Removed FunctionalChaosSobolIndices.getSobolGrouped(Total)Index(int)
* Removed FunctionalChaosSobolIndices::summary
* Removed CorrelationAnalysis::PearsonCorrelation|SpearmanCorrelation|PCC|PRCC
* Removed Distribution.getStandardMoment
* Removed deprecated LinearModelAlgorithm | LinearModelStepwiseAlgorithm ctors
* Removed FunctionalChaosAlgorithm(Function) ctors
* Removed MetaModelResult(Function, Function, ...) ctor
* Removed FunctionalChaosResult.getComposedModel, MetaModelResult.getModel and associated attributes for 1.21
* Removed FunctionalChaosResult Function ctor
* Removed (Process)Sample::computeCenteredMoment
* Removed Distribution::getCenteredMoment
* Removed IterativeMoments::getCenteredMoments
* Removed GeneralLinearModelAlgorithm | KrigingAlgorithm ctors
* Removed Drawable.draw|clean & Graph.draw|clean|getRCommand|makeR*|GetExtensionMap methods
* Removed Sample.storeToTemporaryFile|streamToRFormat methods
* Removed GaussianProcess.GIBBS
* Deprecated DomainIntersection|Union|DisjunctiveUnion(Domain, Domain) ctors
* KrigingResult.getTrendCoefficients now returns a single Point
* GeneralLinearModelResult.getTrendCoefficients now returns a single Point
* KrigingResult.getBasisCollection was removed in favor of KrigingResult.getBasis which returns a single Basis
* GeneralLinearModelResult.getBasisCollection was removed in favor of GeneralLinearModelResult.getBasis which returns a single Basis
* Deprecated EllipticalDistribution::getInverseCorrelation
* HSICStat arguments updated (covariance matrices instead of data + covariance models)
* Deprecated HSICStat relying on weight matrix
* Deprecated Basis::getDimension, use getInputDimension instead
* Removed thinning from Gibbs and all MetropolisHastings classes
* Removed burn-in from Gibbs and all MetropolisHastings classes except RandomWalkMetropolisHastings
* RandomWalkMetropolisHastings::getRealization|Sample no longer remove states reached during burn-in
* RandomWalkMetropolisHastings default adaptation parameters were changed to enable adaptation
* Split BoxCoxFactory::build into buildWithGraph, buildWithLM, buildWithGLM

=== Documentation ===
* Add API documentation to common use cases pages
* Added new use case: Linthurst

=== Python module ===

=== Miscellaneous ===
* Add HypothesisTest::LikelihoodRatioTest for nested model selection
* Add VisualTest::DrawPPplot
* Add VisualTest.Draw(Upper|Lower)(Tail|Extremal)DependenceFunction methods to plot dependence functions
* Add Distribution.compute(Upper|Lower)(Tail|Extremal)DependenceMatrix methods to compute dependence coefficients
* Enable Pagmo.moead_gen with pagmo>=2.19
* Enable Bonmin.Ecp/iFP algorithms with bonmin>=1.8.9
* BoxCoxFactory handles linear model

=== Bug fixes ===
* 2045 ([Debian] NLopt issues during tests on arm64, ppc64el, s390x)
* 2046 ([Debian] DistFunc_binomial issues in tests on arm64 and ppc64el)
* 2047 ([Debian] pythoninstallcheck_DistributionFactory_std issues on arm64, armel, armhf, mips64el)
* 2153 (HSIC computation cost)
* 2185 (Error: no member named '__1' in namespace 'std' ARM64 Android Termux)
* 2193 (Tiny spelling issues)
* 2194 (SubsetSampling: bug in 1.15)
* 2200 (out of bound probas)
* 2204 (Build fails with primesieve-11.0)
* 2205 (Edges of a PolygonArray)
* 2206 (Only Contour legends are shown in a graph mixing Contour and other Drawable (Curve, Cloud,...))
* 2209 (The Felhberg algorithm can fail, sometimes)
* 2210 (Python Domain.getImplementation does not work)
* 2213 (Some script fails)
* 2229 (Unary minus for distribution)
* 2240 (The Viewer sometimes fail)
* 2250 (HSIC Target sensitivity bad filtering)
* 2252 (The examples of calibration are difficult to understand)
* 2255 (The description of a Distribution is wrong)
* 2268 (MeixnerDistribution: slow pdfgradient)
* 2274 (Basis accepts functions with different input (output) dimensions)
* 2281 (The marginal of FireSatelliteModel can fail)
* 2285 (The input and output descriptions does not go down to the metamodel)
* 2287 (DesignProxy.computeDesign() can produce a segmentation fault)
* 2296 (DiracCovarianceModel.discretize() is buggy)
* 2297 (Legend location with grid layout)
* 2299 (Allow openturns as subproject)
* 2306 (Dirichlet::computeConditionalPDF can produce NaNs and Infs)
* 2323 (Minor wording issues in the cross entropy importance sampling example)
* 2327 (A sign mistake in FGM document)
* 2328 (Sample.add weird behaviour)
* 2338 (HSIC with large amount of data makes crash)

=== Library ===

==== New classes ====
* IndependentCopulaFactory
* FieldToPointFunctionalChaosAlgorithm (openturns.experimental)
* FieldFunctionalChaosResult (openturns.experimental)
* FieldFunctionalChaosSobolIndices (openturns.experimental)
* CorrelationAnalysis
* UserDefinedMetropolisHastings (openturns.experimental)
* QuantileMatchingFactory
* UniformMuSigma

==== API changes ====
* Removed coupling_tools.execute get_stderr,get_stdout arguments
* Removed deprecated Nlopt|Ceres|Bonmin|CMinpack|Ipopt|TBB|HMatrixFactory::IsAvailable methods
* Deprecated SequentialStrategy
* Deprecated TensorApproximationAlgorithm, TensorApproximationResult, CanonicalTensorEvaluation, CanonicalTensorGradient
* Deprecated FunctionalChaosSobolIndices.getSobolGrouped(Total)Index(int)
* Deprecated static CorrelationAnalysis::PearsonCorrelation|SpearmanCorrelation|PCC|PRCC, use the methods of the new CorrelationAnalysis class instead.
* Removed static CorrelationAnalysis::SRC|SRRC due to a bug: 1753.
* Deprecated Distribution::getStandardMoment
* Inverted ctors arguments order in LinearModelAlgorithm & LinearModelStepwiseAlgorithm : first sample, then basis
* Deprecated oldest LinearModelAlgorithm(X, basis, Y) and LinearModelStepwiseAlgorithm(X, basis, Y,...) ctors
* Deprecated FunctionalChaos(Function, ...) ctors
* Deprecated MetaModelResult(Function, Function, ...) ctor
* Deprecated FunctionalChaosResult.getComposedModel, MetaModelResult.getModel
* Deprecated FunctionalChaosResult(Function, ...) ctor
* Deprecated (Process)Sample::computeCenteredMoment in favor of computeCentralMoment
* Deprecated Distribution::getCenteredMoment in favor of getCentralMoment
* Deprecated IterativeMoments::getCenteredMoments in favor of getCentralMoments
* Deprecated GeneralLinearModelAlgorithm | KrigingAlgorithm ctors using collection of basis
* Deprecated Drawable.draw|clean & Graph.draw|clean|getRCommand methods for legacy R graphs
* Deprecated SampleImplementation.storeToTemporaryFile|streamToRFormat
* Deprecated GaussianProcess.GIBBS in favor of GaussianProcess.GALLIGAOGIBBS
* Deprecated FunctionalChaosSobolIndices::summary in favor of __str__
* Deprecated BoxCoxFactory::build method using collection of basis

=== Documentation ===
* Added example galleries at the end of API doc pages
* Added new use case: WingWeightModel and example of use
* Added new use case: FireSatelliteModel and example of use

=== Python module ===
* New openturns.experimental submodule introducing newest classes until stabilization

=== Miscellaneous ===
* Chaos for mixed variables

=== Bug fixes ===
* 1214 (There is no example with IntegrationStrategy on a database)
* 1333 (Polynomial chaos with mixed variables (improvement))
* 1473 (FunctionalChaosResult should provide the used input/output samples )
* 1568 (There is no example to bootstrap the polynomial chaos)
* 2030 (There is no example which shows how to calibrate a model without observed inputs)
* 2043 (Some attributes of PythonDistribution are changed during the lifecycle of the object)
* 2058 (TruncatedNormal failure)
* 2059 (Pb with computeConditionalPDF in KernelMixture)
* 2064 (drop getGroupedSobolIndex(int) ?)
* 2073 (getCenteredMoment(0))
* 2076 (Problem in TruncatedDistribution of discrete distributions)
* 2083 (Problem in the QQplot of a discrete distribution)
* 2088 (Why is this library overriding the SIGINT handler?)
* 2089 (Cannot load Python objects with large attributes in a Study)
* 2097 (LinearModelStepwiseAlgorithm ctor order)
* 2091 (Compressed H5 files?)
* 2094 (getSampleAtVertex() is not robust)
* 2095 (Triangular::computeCharacteristicfunction() produces NaNs)
* 2098 (LinearModelStepwiseAlgorithm null basis)
* 2101 (drop FunctionalChaosSobolIndices::summary)
* 2103 (FunctionalChaosRandomVector notes)
* 2110 (HSIC draw indices method does not handle input sample description names)
* 2115 (Add mini-galleries of examples on the API pages)
* 2121 (GaussianProcess Gibbs sampling method issues)
* 2123 (MixtureClassifier 'grade' method does not work with a Sample input)
* 2125 (SciPyDistribution __init__ failure with scipy 1.9.0)
* 2129 (get samples from Wishart distribution)
* 2139 (KarhunenLoeveSVDAlgorithm seems to truncate the expansion from v1.19)
* 2140 (GeneralizedParetoFactory.buildMethodOfMoments estimating wrong parameter)
* 2145 (Calibration wo obs input API)
* 2152 (P1LagrangeInterpolation can make Python fail)
* 2154 (SpaceFillingMinDist: LaTeX typo)
* 2157 (Slow creation of mixtures when the atom distributions are costly to copy)
* 2161 (Drop Normal SPD check)
* 2176 (Missing documentation on Kriging Result methods)