Math.NET Numerics Release Notes
Math.NET Numerics  MKL Provider  OpenBLAS Provider  Data Extensions
4.8.1  20190612
 Packaging fix  signed edition was not properly strong named.
4.8.0  20190602
 Optimization: LevenbergMarquardt, TrustRegion Dogleg ~Jong Hyun Kim
 Optimization: NelderMeadSimplex: Improve convergence for symmetrical functions ~Erik Ovegard
 BUG: Optimization: NelderMeadSimplex did not return the best evaluated point in some cases ~Eric Scott Salem
 Factorial: first 170 values now constant (data segment) instead of precomputed on first use ~Portalez Regis
 Window Functions: Tukey window ~Marco Ross
4.7.0  20181111
 Special Functions: Airy functions Ai, Bi ~Jong Hyun Kim
 Special Functions: Bessel functions of the first and second kind ~Jong Hyun Kim
 Special Functions: Modified Bessel functions of the first and second kind ~Jong Hyun Kim
 Special Functions: Spherical Bessel functions of the first and second kind ~Jong Hyun Kim
 Special Functions: Hankel functions of the first and second kind ~Jong Hyun Kim
 Special Functions: Kelvin functions of the first and second kind, and derivatives ~Jong Hyun Kim
 Linear Algebra: optimized sparse implementation of transposemultiply ~Richard Reader
 Linear Algebra: optimized range checking in vectors and matrices
4.6.0  20181019
 Polynomial: New Polynomial data structure ~Tobias Glaubach
 Statistics: Correlate.Auto auto correlation ~Tobias Glaubach
 Curve Fitting: Fit.Curve for nonlinear leastsquares fitting to an arbitray function
4.5.1  20180522
 Signed NuGet packages
4.5.0  20180522
 Random: New Xoshiro256StarStar random source ~Colin Green
4.4.1  20180506
 Build: strongname edition .Signed NuGet packages are available again
4.4.0  20180225
 Linear Algebra: managed provider performance improvements on basic vector arithmetics
4.3.0  20180224
 FFT: migrate managed algorithms to provider, streamline implementation wiring, scaling logic.
 FFT: obsoleted algorithmspecific functions like Radix2Forward, they use Forward internally until v5.
 FFT: migrate "naive" implementation to tests as reference implementation.
 Build: fix NuGet packaging sucht that the included assemblies are really codesigned.
4.2.0  20180221
 Native Providers: MklProvider and similar types now public (entry point for memory management, etc)
 Native Providers: All providers now support freeing resources without unloading the provider
 Native Providers: MKL provider sets consistency, precision and accuracy modes earlier to ensure they are applied
 Native Providers: If a provider has been loaded successfully, skip any future loading attempts (faster switching)
 Build: add .Net Framework 4.6.1 target (main package), switch to 4.6.1 for testing projects
4.1.0  20180219
 Curve Fitting: Fit.Power, Fit.Exponential, Fit.Logarithm (similar to excel trend curves)
4.0.0  20180211
 Explicit builds for .Net Framework 4.0 and higher and for .Net Standard 1.3 and 2.0
 F#: explicit builds for .Net Framework 4.5 and higher and for .Net Standard 1.6 and 2.0
 BREAKING: drop obsolete functionality (which has been marked as obsolete for a while)
 BREAKING: drop support for .Net 3.5 and PCL profiles
 BREAKING: F#: drop support for .Net 4.0 and PCL profiles
 BREAKING: F#: depends on FSharp.Core 4.3.3
 BREAKING: Native provider implementation types internal (but can be constructed explicitly)
 BREAKING: Native provider control API moved from Control to perprovider Control classes
 BREAKING: Linear Algebra: Vector.Map consistent with Matrix.Map, automatic fallback to inplace
 BREAKING: Linear Algebra: Storage providers must always force all parameters (no defaults)
 BREAKING: Statistics: Ranks and RankInplace of single precision now return single precision arrays
 BUG: Ode Solver: fix typo in the RungeKutta solvers on timestep handling ~Ksero
 BUG: fix Matrix.GetHashCode for wide matrices ~mjmckp
 BUG: Distributions: BetaScaled no longer ignores optional random source parameter
 BUG: Trigonometry: Complex trigonometric functions behave better on large imput ~diluculo
 BUG: Trigonometry: Fix imaginary part sign of complex hyperbolic cotangent
 Control.Describe: human readable summary of the effective Math.NET Numerics configuration
 Providers: Native Provider support in .Net Standard 2.0 (but not in 1.x).
 Linear Algebra: F# vector/matrix functions to accept all #seq instead of lists only
 Linear Algebra: Vector MapInplace implemented at storage level
 Curve Fitting: Fit.LineThroughOrigin shortcut
 Optimization: LimitedMemory BFGS ~Florian Wechsung
 Root Finding: Broyden: step size for calculating appox Jacobian, more robust step size formula ~Aappo Pulkkinen
 Statistics: Kernel Density Estimation ~Christoph Albert
 Statistics: GoodnessOfFit CoefficientOfDetermination ~Jon Smit
 Fixed a whole range of inline documentation typos, misc code quality fixes ~Jonas Nyrup
 Examples: moved to examples folder, new framework target integration project file examples
 Benchmarks: builds for both net46 and netcoreapp2.0
 Assemblies are now codesigned with an X.509 certificate, using SHA2256 hashes
4.0.0beta07  20180210
 Curve Fitting: Fit.LineThroughOrigin shortcut
 Build: Target FSharp.Core 4.3.3 (for better netstandard2.0 compatibility)
 Build: use latest SDKs and runtimes, cleanups
4.0.0beta06  20180203
 Rework conditional compilation to simplify and increase the .Net Standard api surface
 BREAKING: Native provider implementation types internal (but can be constructed explicitly)
 BREAKING: Native provider control API moved from Control to perprovider Control classes
 Control.Describe: human readable summary of the effective Math.NET Numerics configuration
 Fixed a whole range of inline documentation typos ~Jonas Nyrup
 Build: reworked test build targets, switched to Paket magic mode
 Examples: moved to examples folder, new framework target integration project file examples
4.0.0beta05  20180122
 BUG: Distributions: BetaScaled no longer ignores optional random source parameter
 BUG: Trigonometry: Complex trigonometric functions behave on large imput ~diluculo
 All: code quality fixes ~Jonas Nyrup
4.0.0beta04  20180113
 BREAKING: Linear Algebra: Vector.Map consistent with Matrix.Map, automatic fallback to inplace
 BREAKING: Linear Algebra: Storage providers must always force all parameters (no defaults)
 Linear Algebra: F# vector/matrix functions to accept all #seq instead of lists only
 Linear Algebra: Vector MapInplace implemented at storage level
4.0.0beta03  20180109
 Optimization: LimitedMemory BFGS ~Florian Wechsung
 BUG: Trigonometry: Fix imaginary part sign of complex hyperbolic cotangent
4.0.0beta02  20180107
 Root Finding: Broyden: step size for calculating appox Jacobian, more robust step size formula ~Aappo Pulkkinen
 Statistics: Kernel Density Estimation ~Christoph Albert
4.0.0beta01  20180107
 BREAKING: drop obsolete functionality (which has been marked as obsolete for a while)
 BREAKING: drop support for .Net 3.5 and PCL profiles
 BREAKING: F#: drop support for .Net 4.0 and PCL profiles
 BREAKING: F#: depends on FSharp.Core 4.2.3
 Explicit builds for .Net Framework 4.0 and higher and for .Net Standard 1.3 and 2.0
 F#: explicit builds for .Net Framework 4.5 and higher and for .Net Standard 1.6 and 2.0
 Assemblies are now codesigned with an X.509 certificate, using SHA2256 hashes
 BUG: Ode Solver: fix typo in the RungeKutta solvers on timestep handling ~Ksero
 BUG: fix Matrix.GetHashCode for wide matrices ~mjmckp
3.20.2  20180122
 Bugfixes backported from v4:
 BUG: Distributions: BetaScaled no longer ignores optional random source parameter
 BUG: Trigonometry: Complex trigonometric functions behave on large imput ~diluculo
3.20.1  20180113
 Bugfixes backported from v4:
 BUG: Trigonometry: Fix imaginary part sign of complex hyperbolic cotangent
 BUG: Ode Solver: fix typo in the RungeKutta solvers on timestep handling ~Ksero
 BUG: fix Matrix.GetHashCode for wide matrices ~mjmckp
3.20.0  20170715
 Optimization: nonlinear optimization algorithms ~Scott Stephens, Erik Ovegard, bdodson, et al.
 Native Providers: from now on also supported in the .Net 3.5 build.
 BUG: Special Functions: allow more iterations in BetaRegularized ~Elias Abou Jaoude
3.19.0  20170429
 Statistics: RunningStatistics.Combine to better handle empty statistics ~Lucas Godshalk
 Linear Algebra: Cholesky.Factorize to reuse the factorization matrix ~mjmckp
 Linear Algebra: Fix docs for DoPointwiseMultiply ~Jakub Arnold
3.18.0  20170409
 FFT: singleprecision support ~AlexHild
3.17.0  20170115
 Random: random sources (all except crypto) now support ephemeral serialization.
 Linear Algebra: explicit impl to copy a range of a row of a sparse matrix to a range of a sparse vector ~arthurvb
 Linear Algebra: explicitly demand fully modifiable matrix where needed, fixes issues with diagonal matrices.
 FFT: leverage new matrix internal array access approach in 2D matrix transformations.
3.16.0  20170103
 Root Finding: improve accuracy handling ~Konstantin Tretyakov
 Regression: GoodnessOfFit StandardError ~David Falkner
3.15.0  20161227
 FFT: MKL native provider backend.
 FFT: 2D and multidimensional FFT (only supported by MKL provider, managed provider pending).
 FFT: real conjugateeven FFT (only leveraging symmetry in MKL provider).
 FFT: managed provider significantly faster on x64.
 Linear Algebra: pointwise trigonometric and basic functions ~Albert Pang
 Linear Algebra: better support for F# builtin operators (sqrt, sin, exp, ..) ~Albert Pang
 Linear Algebra: pointwise power operator (F#)
 Linear Algebra: enable experimental matrix product implementation
 Linear Algebra: better support for matrix to/from rowmajor arrays and enumerables
 Linear Algebra: transport allows specifying a result matrix to transpose into, inplace if square
 Linear Algebra: vector and matrix AsArray and similar to access internal arrays if applicable
 Linear Algebra: vector and matrix pointwise min/max and absmin/absmax
 Linear Algebra: dotpower on vectors and matrices, supporting native providers.
 Linear Algebra: matrix MoorePenrose pseudoinverse (SVD backed).
 Provider Control: separate Control classes for LA and FFT Providers.
 Provider Control: avoid internal exceptions on provider discovery.
 Distributions: fix misleading inline docs on NegativeBinomial.
 Generate: linear integer ranges
 Root Finding: extend zerocrossing bracketing in derivativefree algorithms.
 Window: periodic versions of Hamming, Hann, Cosine and Lanczos windows.
 Special Functions: more robust GammaLowerRegularizedInv (and Gamma.InvCDF).
 BUG: ODE Solver: fix bug in RungeKutta second order routine ~Ksero
3.13.1  20160906
 BUG: Random: Next(x,x+1) must always return x ~Juri
3.13.0  20160818
 Linear Algebra: faster tall, wide managed matrix multiplication. ~Aixile
 Euclid: Integer Log2 (DeBruijn sequencences algorithm).
 Integration: GaussLegendre documentation, cleanup. ~Larz White
 Random: Integer subrange sampling to use rejection sampling to avoid bias.
 Random: Improvements on integer and byte sampling.
 BUG: Random: CryptoRandomSource must not generate 1.0 samples.
 BUG: Statistics: fixed bug in WeightedPearson Correlation. ~Jon Smit
3.12.0  20160703
 ODE Solver: RungeKutta (order 2, 4) and AdamsBashforth (order 14) algorithms ~Yoonku Hwang
 Linear Algebra: faster multiplication of sparse with dense matrices ~Arthur
 BUG: Integration: GaussLegendre on order 256 ~Sergey Kosukhin
 BUG: Distributions: ChiSquared sampling was taking a square root where it should not ~Florian Wechsung
3.11.1  20160424
 BUG: Linear Algebra: sparse vector pointwise multiply/divide to itself
 BUG: Linear Algebra: Vector.ToVectorString if the first column is wider than maxWidth
3.11.0  20160213
 Special Functions: error functions to use static coefficient arrays (perf) ~Joel Sleppy
 Integration: GaussLegendre Rule (1D, 2D) ~Larz White
 Complex: more robust magnitude and division for numbers close to MaxValue or Epsilon ~MaLiN2223
 Native Providers: lazy default provider discovery & initialization ~Kuan Bartel
 FSharp Package: Quaternion type ~Phil Cleveland
3.10.0  20151230
 Statistics: singleprecision floating point support.
 Statistics: very limited support for int32 and complex numbers.
 Statistics: Min/Max Absolute, MagnitudePhase (complex).
 Statistics: FiveNumberSummary to use actual Median instead of R8 quantile.
 Linear Algebra: matrix Rank to use relative epsilon.
 Linear Algebra: extensions to convert between single/double precision, complex/real.
 Linear Algebra: Vector/Matrix storage DataContracts for ephemeral serialization.
 Regression: more helpful exceptions and messages.
 Random: 'Next' integer sampling no longer involves floating points, avoids oneoff error in MersenneTwister.
 Precision: EpsilonOf for singleprecision numbers, drop no longer needed portable fallbacks.
3.9.0  20151125
 Distributions: Normal.CDF avoids problematic subtraction by using Erfc instead of Erf.
 Statistics: geometric and harmonic mean.
 Statistics: DataContracts for ephemeral serialization on RunningStatistics, DescriptiveStatistics and Histogram.
 BUG: Statistics: Histogram did not adjust lower bound correctly when value was equal to the bound ~Volker Breuer
 Linear Algebra: minor optimization on how we call Array.Copy.
 BUG: Linear Algebra: fix bug in Complex and Complex32 SparseMatrix.ToTypeString.
3.8.0  20150926
 Distributions: PDF and CDF more robust for large distribution parameters.
 Distributions: BetaScaled distribution.
 Distributions: method to create a PERT distribution (based on BetaScaled) ~John C Barstow
 Distributions: Weibull.Estimate ~Jon Larborn
 Random: NextBoolean extensions.
 Root Finding: RootFinding.Secant (based on NewtonRaphson) ~grovesNL
 Linear Algebra: Matrix Rank calculation now uses a tolerance based on the matrix size.
 Linear Algebra: Alternative CreateMatrix/Vector functions with type parameter on functions instead of type.
 Linear Algebra: MKL LinearAlgebra provider requires at least native provider r9 (linear algebra v2.0).
 Native Providers: automatic handling of intermediate work arrays/buffers in MKL and OpenBLAS providers ~Marcus Cuda, Kuan Bartel
 Native Providers: automatically use native provider if available.
 Native Providers: new Control.TryUse* to make it simpler to use providers if available but without failing if not.
 Native Providers: improved error state checking and handling ~Marcus Cuda, Kuan Bartel
 Combinatorics: generate or select random permutation, combination or variation (shuffling)
 Finance: rename CompoundMonthlyReturn to CompoundReturn (old now obsolete).
3.7.1  20150910
 BUG: Linear Algebra: fix optimized path of adding a sparse matrix to itself.
3.7.0  20150509
 Statistics: RunningStatistics now propagates min/max on Combine, handles NaN on Push.
 Statistics: new MovingStatistics providing descriptive statistics over a moving window ~Marcus Cuda
 Statistics: new Statistics.MovingAverage.
 Statistics: Improved Histogram handling of smallwidth buckets ~Justin Needham
 Distributions: ChiSquare.InvCDF ~logophobia
 FFT: Fourier.FrequencyScale to generate the frequency corresponding to each index in frequency space.
 BUG: FFT: fix Bluestein algorithm for sequences with more than 46341 samples but not poweroftwo.
 Linear Algebra: SparseVector.AbsoluteMaximumIndex ~Matt Heffron
 MKL Native Provider: OSX build script ~Marcus Cuda
 MKL Native Provider: new combined NuGet package with a proper build target (no more manual file handling needed).
 OpenBLAS Native Provider: a new linear algebra provider using OpenBLAS ~Kuan Bartel
 CUDA Native Provider: a new GPUbased linear algebra provider using Nvidia CUDA ~Matthew A. Johnson
 Native Providers: now versioned separately for each kind (MKL, CUDA, OpenBLAS).
3.6.0  20150322
 Distributions: ChiSquare.CDF more robust for large numbers ~Baltazar Bieniek
 Linear Algebra: MatrixStorage.Map2 equivalent to VectorStorage.Map2
 Linear Algebra: Matrix and Vector Find/Find2, Exists/Exists2, ForAll/ForAll2
 Linear Algebra: more consistent range checking in MatrixStorage.Clear and related
 Linear Algebra: mixedstorage fall back implementations now leverage higherorder functions
 BUG: Linear Algebra: fix loop range in MatrixStorage.ClearColumns (builtin storage not affected)
 BUG: Linear Algebra: fix sparse matrix equality.
 BUG: Linear Algebra: ArgumentException instead of index exception when trying to create an empty matrix.
 Generate: Unfold, Fibonacci; Normal and Standard replacing Gaussian and Stable.
 Native Providers: NativeProviderLoader to automatically load the provider for the matching processor architecture (x86, x64) ~Kuan Bartel
 Native Providers: Control.NativeProviderPath allowing to explicitly declare where to load binaries from.
 MKL Native Provider: support for native complex eigenvalue decomposition ~Marcus Cuda
 MKL Native Provider: nonconvergence checks in singularvalue and eigenvalue decompositions ~Marcus Cuda
3.5.0  20150110
 Differentiation: derivative, partial and mixed partial; hessian & jacobian ~Hythem Sidky
 Differentiation: Differentiate facade class for simple use cases
 Differentiation: F# module for better F# function support.
 Linear Algebra: matrix ToRowArrays/ToColumnArrays
 Linear Algebra: F# insertRow, appendRow, prependRow and same also for columns
 Linear Algebra: F# append, stack and ofMatrixList2
 Precision: measured machine epsilon, positive vs negative epsilon
3.4.0  20150104
 Special Functions: Generalized Exponential Integral ~Ashley Messer
 Special Functions: Regularized Incomplete Gamma domain extended to a=0 ~Ashley Messer
 Statistics: weighted Pearson correlation ~ViK
 MKL Native Provider: memory functions to free buffers and gather usage statistics ~Marcus Cuda
 F#: depend on new official FSharp.Core NuGet package instead of FSharp.Core.Microsoft.Signed
 F#: simpler NuGet package dependencies (no more need for framework groups)
 Build: vagrant bootstrap now uses the latest xamarin mono packages
3.3.0  20141126
 Linear Algebra: Vector.Fold2 (fold2 in F#), storage optimized
 Linear Algebra: Minor change how matrix products call the LA provider
 Linear Algebra: Random generation now leveraging array sampling routines
 BUG: Linear Algebra: fix bug when manually assigning System.Random to random distribution
 Root Finding: Change Brent tolerance check, add bracket check ~Hythen Sidky
 Root Finding: Auto zerocrossing bracketing in FindRoots facade (not in algorithms)
 Statistics: RootMeanSquare (RMS)
 Distributions: Array sampling routines now available through interface
 Distributions: Categorical sampling now explicitly skips p=0 categories
 Generate: leverage array sampling routines for random data generation
 Generate: square, triangle and sawtooth waves
 Distance: Jaccard Index
 F#: explicitly depend on official FSharp.Core NuGet packages
 F#: NuGet package with iPython IfSharp F# module integration load script
 F#: load scripts with better packet support (and NuGet with ExcludeVersion)
 Build: unified build.sh and buildn.sh into combined build.sh
 Build: use Paket instead of NuGet to maintain NuGet dependencies
 Build: for core add PCL profiles 7, 78 and 259; for F# extensions drop PCL profile 328
3.2.3  20140906
 BUG: MatrixNormal distribution: fix density for nonsquare matrices ~Evelina Gabasova
3.2.2  20140905
 BUG: MatrixNormal distribution: density computation switched row and column covariance ~Evelina Gabasova
3.2.1  20140805
 Package fix: make sure .Net 3.5only dependencies are not installed on .Net 4 and newer.
3.2.0  20140805
 Linear Algebra: Vector.Map2 (map2 in F#), storageoptimized
 Linear Algebra: fix RemoveColumn/Row early index bound check (was not strict enough)
 Statistics: Entropy ~Jeff Mastry
 Interpolation: use Array.BinarySearch instead of local implementation ~Candy Chiu
 Resources: fix a corrupted exception message string
 Portable Build: support .Net 4.0 as well by using profile 328 instead of 344.
 .Net 3.5: F# extensions now support .Net 3.5 as well
 .Net 3.5: NuGet package now contains proper 3.5only TPL package dependency; also in Zip package
3.1.0  20140720
 Random: generate a sequence of integers within a range in one go
 Distributions: all distributions must have static routines to sample an array in one go
 Linear Algebra: fix Matrix.StrictlyLowerTriangle
 Linear Algebra: fix vector DoOuterProduct ~mjmckp
 Linear Algebra: enumerators accept Zerosparameter (like map/fold already does)
 Linear Algebra: Vector.MapConvert (consistency)
 Linear Algebra: proper term for 'conjugate symmetric' is 'Hermitian'
 Interpolation: new Step, LogLinear and transformed interpolators ~Candy Chiu
 Interpolation: check for min required number of data points, throw ArgumentException if not.
 Root Finding: F# FindRoots.broyden module function ~teramonagi
 Misc docs fixes
3.0.2  20140626
 BUG: fixing a bug in Matrix.RemoveRow range checks.
3.0.1  20140624
 BUG: fixing a bug in new Matrix.ToMatrixString and Vector.ToVectorString routines.
3.0.0  20140621

First stable v3 release:
 Upgrade Notes
 Stable API, no more breaking changes for all future v3 releases (except previews).
 Finally unlocks development and contributions around nonlinear optimization and native providers over the next few minor releases.
 Native Providers: option to control max number of threads used by MKL.
 F#: Fit.multiDim; Matrix.qr, svd, eigen, lu and cholesky.
3.0.0beta05  20140620
 2nd Candidate for v3.0 Release
 BUG: Distance: fix bug in Hamming distance that skipped the first pair.
 F#: packages now include a MathNet.Numerics.fsx script that includes FSI printers and references the assemblies.
 Linear Algebra: improved matrix and vector ToString formatting, more compact, adaptive to actual numbers.
 Linear Algebra: CoerceZero for matrix and vector to replace small numbers with zero.
 Regression: DirectRegressionMethod option to specify as argument which direct method should be used.
 Control: drop MaxToStringRows/Columns properties (no longer used)
 Random: clarify bad randomness properties of SystemRandomSource.FastDoubles (trade off)
3.0.0beta04  20140616
 Candidate for v3.0 Release

Linear Algebra:
 FoldRows renamed to FoldByRow, now operates on and returns arrays; same for columns. Breaking.
 New FoldRows and ReduceRows that operate on row vectors; same for columns
 Split Map into Map and MapConvert (allows optimization in common inplace case)
 Row and column sums and absolutesums
 F# DiagonalMatrix module to create diagonal matrices without using the builder
 F# Matrix module extended with sumRows, sumAbsRows, normRows; same for columns
 Build: extend build and release automation, automatic releases also for data extensions and native providers
3.0.0beta03  20140605
 Linear Algebra: vector outer product now follows common style, supports explicit result argument, more efficient.
 Interpolation: must not modify/sort original data; alternative Sorted and Inplace functions.
 Distributions: static IsValidParameterSet functions.
 Distributions: all distributions are now immutable in their distribution parameters. Breaking.
 NuGet: attempt to create proper symbol+source packages on symbolsource; primary packages smaller, w/o pdbs
 Build: skip long tests with new "quick" argument (FAKE)
 Build: clearing is more explicit, fixes most locking issues if solution is also open in IDE.
 Build: automated publishing docs, api, git release tag (maintainer)
3.0.0beta02  20140529

Linear Algebra:
 optimized sparsesparse and sparsediagonal matrix products. ~Christian Woltering
 transpose at storage level, optimized sparse transpose. ~Christian Woltering
 optimized inplacemap, indexed submatrixmap.
 optimized clearing a set of rows or columns.
 matrix FoldRows/FoldColumns.
 matrix column/row norms, normalization.
 prefer enums over boolean parameters (e.g.
Zeros.AllowSkip
).  IsSymmetric is now a method, add IsConjugateSymmetric. Breaking.
 Eigenvalue decomposition can optionally skip symmetry test.
 Direct diagonalscalar division implementation
 Test Functions: Rosenbrock, Rastrigin, DropWave, Ackley, Bohachevsky, Matyas, SixHumpCamel, Himmelblau
 Statistics: DescriptiveStatistics support for larger datasets.
 MKL: native providers must not require MFC to compile.
 Sorting helpers support subrange sorting, use insertion sort on very small sets.
 Build: extend usage of FAKE, automate docs, api, Zip and NuGet package generation.
 Portable: replace PCL profile 136 with profile 344, support for WP8.1
 Xamarin: prepare for better Xamarin Android/iOS support and for adding to the Xamarin store (free).
 Misc code style fixes.
 Update Vagrant setup to official Ubuntu 14.04 LTS box and proper aptstyle Mono+F# provisioning.
3.0.0beta01  20140401
 See also: Roadmap and Towards Math.NET Numerics Version 3.
 Major release with breaking changes
 All obsolete code has been removed
 Reworked redundancies, inconsistencies and unfortunate past design choices.
 Significant namespace simplifications (30%).

Linear Algebra:
 Favor and optimize for generic types, e.g.
Vector<double>
.  Drop the
.Generic
in the namespaces and flattened solver namespaces.  F#: all functions in the modules now fully generic, including the
matrix
function.  F#:
SkipZeros
instead of the crypticnz
suffix for clarity.  Add missing scalarmatrix routines.
 Optimized mixed densediagonal and diagonaldense operations (500x faster on 250k set).
 More reasonable choice of return structure on mixed operations (e.g. dense+diagonal).
 Add pointwise infix operators
.*
,./
,.%
where supported (F#)  Vectors explicitly provide proper L1, L2 and Linfinity norms.
 All norms return the result as double (instead of the specific value type of the matrix/vector).
 Matrix Linfinity norm now cacheoptimized (810x faster).
 Vectors have a
ConjugateDotProduct
in addition toDotProduct
. Matrix.ConjugateTransposeAndMultiply
and variants. Matrix Factorization types fully generic, easily accessed by new
Matrix<T>
member methods (replacing the extension methods). Discrete implementations no longer visible.  QR factorization is thin by default.
 Matrix factorizations no longer clone their results at point of access.
 Add direct factorizationbased
Solve
methods to matrix type.  Massive iterative solver implementation/design simplification, now mostly generic and a bit more functionalstyle.
 Renamed iterative solver stop criteria from 'criterium' to 'criterion'.
 New MILU(0) iterative solver preconditioner that is much more efficient and fully leverages sparse data. ~Christian Woltering
 Matrices/Vectors now have more consistent enumerators, with a variant that skips zeros (useful if sparse).
 Matrix/Vector creation routines have been simplified and usually no longer require explicit dimensions. New variants to create diagonal matrices, or such where all fields have the same value. All functions that take a params array now have an overload accepting an enumerable (e.g.
OfColumnVectors
).  Generic Matrix/Vector creation using builders, e.g.
Matrix<double>.Build.DenseOfEnumerable(...)
 Create a matrix from a 2Darray of matrices (topleft aligned within the grid).
 Create a matrix or vector with the same structural type as an example (
.Build.SameAs(...)
)  Removed nonstatic Matrix/Vector.CreateMatrix/CreateVector routines (no longer needed)
 Add Vector.OfArray (copying the array, consistent with Matrix.OfArray  you can still use the dense vector constructor if you want to use the array directly without copying).
 More convenient and one more powerful overload of
Matrix.SetSubMatrix
.  Matrices/Vectors expose whether storage is dense with a new IsDense property.
 Various minor performance work.
 Matrix.ClearSubMatrix no longer throws on 0 or negative col/row count (nop)
 BUG: Fix bug in routine to copy a vector into a subrow of a matrix.
 Both canonical modulus and remainder operations on matrices and vectors.
 Matrix kernel (null space) and range (column space)
 Storageaware noninplace functional map on vectors and matrices
 Pointwise power, exponential and natural logarithm for vectors and matrices.
 Matrix positiveinteger power
 Matrix RemoveRow/RemoveColumn; more efficient InsertRow/InsertColumn
 Favor and optimize for generic types, e.g.

Native Linear Algebra/Intel MKL:
 Thin QR factorization uses MKL if enabled for all types (previously just
double
)  Sparse matrix CSR storage format now uses the much more common row pointer convention and is fully compatible with MKL (so there is nothing in the way to add native provider support).
 Providers have been moved to a
Providers
namespace and are fully generic again.  Simpler provider usage:
Control.UseNativeMKL()
,Control.UseManaged()
.  MKL native provider now supports capability querying (so we can extend it much more reliably without breaking your code).
 MKL native provider consistency, precision and accuracy now configurable (tradeoff).
 Native Provider development has been reintegrated into the main repository; we can now directly run all unit tests against local native provider builds. Covered by FAKE builds.
 Thin QR factorization uses MKL if enabled for all types (previously just

Statistics:
 Pearson and Spearman correlation matrix of a set of arrays.
 Spearman ranked correlation optimized (4x faster on 100k set)
 Skewness and PopulationSkewness; Kurtosis and PopulationKurtosis.
 Singlepass
MeanVariance
andMeanStandardDeviation
methods (often used together).  Some overloads for singleprecision values.
 Add
Ranks
,QuantileRank
andEmpiricalCDF
.  F# module for higher order functions.
 Median direct implementation (instead of R8compatible 0.5quantile)
 New RunningStatistics that can be updated and merged
 BUG: DescriptiveStatistics must return NaN if not enough data for a specific statistic.

Probability Distributions:
 Direct static distributions functions (PDF, CDF, sometimes also InvCDF).
 Direct static sample functions, including such to fill an existing array in one call.
 New Trigangular distribution ~Superbest, David Prince
 Add InvCDF to Gamma, StudentT, FisherSnedecor (F), and Beta distributions.
 Major API cleanup, including xml docs
 Xml doc and ToString now use wellknown symbols for the parameters.
 Maximumlikelihood parameter estimation for a couple distributions.
 All constructors now optionally accept a random source as last argument.
 Use less problematic RNGseeds by default, if no random source is provided.
 Simpler and more composable random sampling from distributions.
 Much more distribution's actual sample distribution is verified in tests (all continuous, most discrete).
 Binomial.CDF now properly leverages BetaRegularized.
 BUG: Fix hypergeometric CDF semantics, clarify distribution parameters.
 BUG: Fix Zipf CDF at x=1.
 BUG: Fix Geometric distribution sampling.
 BUG: Fix Categorical distribution properties. ~David Prince

Random Numbers:
 All RNGs provide static Sample(values) functions to fill an existing array.
 Threadsafe System.Random available again as
SystemRandomSource
.  Fast and simple to use static
SystemRandomSource.Doubles
routine with lower randomness guarantees.  Shared
SystemRandomSource.Default
andMersenneTwister.Default
instances to skip expensive initialization.  Using threadsafe random source by default in distributions, Generate, linear algebra etc.
 Tests always use seeded RNGs for reproducability.
 F#: direct sampling routines in the
Random
module, also including default and shared instances.

Linear Regression:
 Reworked
Fit
class, supporting more simple scenarios.  New
.LinearRegression
namespace with more options.  Better support for simple regression in multiple dimensions.
 Goodness of Fit: R, RSquared ~Ethar Alali
 Weighted polynomial and multidim fitting.
 Use more efficient LA routines ~Thomas Ibel
 Reworked

Interpolation:
 Return tuples instead of out parameter.
 Reworked splines, drop complicated and limiting inheritance design. More functional approach.
 More efficient implementation for noncubic splines (especially linear spline).
Differentiate2
instead ofDifferentiateAll
. Definite
Integrate(a,b)
in addition to existing indefiniteIntegrate(t)
.  Use more common names in
Interpolate
facade, e.g. "Spline" is a well known name.
 Root Finding: Chebychev polynomial roots.
 Root Finding: Cubic polynomials roots. ~Candy Chiu
 Trig functions: common short names instead of very long names. Add sinc function.
 Excel functions: TDIST, TINV, BETADIST, BETAINV, GAMMADIST, GAMMAINV, NORMDIST, NORMINV, NORMSDIST, NORMSINV QUARTILE, PERCENTILE, PERCENTRANK.
 Special functions: BetaRegularized more robust for large arguments.
 Special functions: new
GammaLowerRegularizedInv
.  New distance functions in
Distance
: euclidean, manhattan, chebychev distance of arrays or generic vectors. SAD, MAE, SSD, MSE metrics. Pearson's, Canberra and Minkowski distance. Hamming distance.  Windows: ported windowing functions from Neodym (Hamming, Hann, Cosine, Lanczos, Gauss, Blackmann, Bartlett, ...)
 BigInteger factorial

Build:
 FAKEbased build (in addition to existing Visual Studio solutions) to clean, build, test, document and package independently of the CI server.
 Finally proper documentation using FSharp.Formatting with sources included in the repository so it is versioned and can be contributed to with pull requests.
 NuGet packages now also include the PCL portable profile 47 (.Net 4.5, Silverlight 5, Windows 8) in addition to the normal .Net 4.0 build and PCL profile 136 (.Net 4.0, WindowsPhone 8, Silverlight 5, Windows 8) as before. Profile 47 uses
System.Numerics
for complex numbers, among others, which is not available in profile 136.  NuGet packages now also include a .Net 3.5 build of the core library.
 IO libraries have been removed, replaced with new
.Data
packages (see list on top).  Alternative strongnamed versions of more NuGet packages (mostly the F# extensions for now), with the
.Signed
suffix.  Reworked solution structure so it works in both Visual Studio 11 (2012) and 12 (2013).
 We can now run the full unit test suite against the portable builds as well.
 Builds should now also work properly on recent Mono on Linux (including F# projects).
 Fixed builds on platforms with case sensitive file systems. ~Gauthier Segay
 Integration: simplification of the doubleexponential transformation api design.
 FFT: converted to static class design and shorter names for simpler usage. Drop now redundant
Transform
class.  Generate: ported synthetic data generation and sampling routines from Neodym (includes all from old Signals namespace). F# module for higher order functions.
 Euclid: modulus vs remainder (also BigInteger), integer theory (includes all from old NumberTheory namespace).
 Complex: common short names for Exp, Ln, Log10, Log.
 Complex: fix issue where a negative zero may flip the sign in special cases (like
Atanh(2)
, where incidentally MATLAB and Mathematica do not agree on the sign either).  Complex: routines to return all two square and three cubic roots of a complex number.
 Complex: More robust complex Asin/Acos for large real numbers.
 Evaluate: routine to evaluate complex polynomials, or real polynomials at a complex point.
 CommonParallel now also supported in .Net 3.5 and portable profiles; TaskScheduler can be replaced with custom implementation ~Thomas Ibel
 F# BigRational type cleaned up and optimized ~Jack Pappas
 F# BigRational IsZero, IsOne, IsInteger, create from fraction.
 F# BigRational Reciprocal, Power operator support (), support for negative integer powers.
 F# functions now use the clearer
Func
suffix instead of justF
if they return a function.  Precision: reworked, now much more consistent. If you use
AlmostEqual
with numbersbetween/ULP semantics, please do review your code to make sure you're still using the expected variant!. If you use the decimalplaces semantics, you may need to decrement the digits argument to get the same behavior as before.  Much less null checks, our code generally only throws
ArgumentNullException
if an unexpected null argument would not have caused an immediateNullReferenceException
.  Cases where
ArgumentOutOfRangeExceptions
where thrown with wrong arguments (i.e. no parameter name) now throwArgumentException
instead.  Tests now have category attributes (to selectively run or skip categories).
2.6.2  20131021
 Patch release, fixing the NuGet package to work better in WindowsPhone 8 projects. Assemblies are not changed.
2.6.1  20130813
 BUG: fixing a bug in
ArrayStatistics.Variance
on arrays longer than 46341 entries.
2.6.0  20130726
 See also: What's New in Math.NET Numerics 2.6
 Linear Curve Fitting: Linear leastsquares fitting (regression) to lines, polynomials and linear combinations of arbitrary functions. Multidimensional fitting. Also works well in F# with the F# extensions.

Root Finding:
 Brent's method. ~Candy Chiu, Alexander Täschner
 Bisection method. ~Scott Stephens, Alexander Täschner
 Broyden's method, for multidimensional functions. ~Alexander Täschner
 NewtonRaphson method.
 Robust NewtonRaphson variant that tries to recover automatically in cases where it would fail or converge too slowly. This modification makes it more robust e.g. in the presence of singularities and less sensitive to the search range/interval.
 All algorithms support a TryFindpattern which returns success instead of throwing an exception.
 Special case for quadratic functions, in the future to be extended e.g. to polynomials.
 Basic bracketing algorithm
 Also works well in F# with the F# extensions.

Linear Algebra:
 Native eigenvalue decomposition (EVD) support with our MKL packages ~Marcus Cuda
 Add missing scalarvector operations (sv, s/v, s%v) ~Thomas Ibel
 Support for new F# 3.1 row/column slicing syntax on matrices
 Matrices learned proper OfColumn/RowVectors, analog also in F#.
 Documentation Fixes ~Robin Neatherway
 BUG: Fixed exception text message when creating a matrix from enumerables (rows vs columns) ~Thomas Ibel
 We're phasing out MathNet.Numerics.IO that used to be included in the main package for matrix file I/O for text and Matlab formats. Use the new .Data.Text and .Data.Matlab packages instead.
 Statistics: Spearman Rank Correlation Coefficient ~Iain McDonald
 Statistics: Covariance function, in Array, Streaming and common Statistics.
 Distributions: Categorical: distribution more consistent, no longer requires normalized pdf/cdf parameters
 Distributions: Categorical: inverse CDF function ~Paul Varkey
 BUG: Distributions: Fixed static sampling methods of the
Stable
distribution. ~Artyom Baranovskiy  BUG: Fixed a bug in the Gamma Regularized special function where in some cases with large values it returned 1 instead of 0 and vice versa.
 The F# extensions now have a strong name in (and only in) the signed package as well (previously had not been signed). ~Gauthier Segay
 Evaluate.Polynomial with new overload which is easier to use.
 Fixed a couple badly designed unit tests that failed on Mono.
 Repository now Vagrantready for easy testing against recent Mono on Debian.
2.5.0  20130414
 See also: What's New in Math.NET Numerics 2.5
 Statistics: Empty statistics now return NaN instead of either 0 or throwing an exception. This may break code in case you relied upon the previous unusual and inconsistent behavior.
 Linear Algebra: More reasonable ToString behavior for matrices and vectors. This may break code if you relied upon ToString to export your full data to text form intended to be parsed again later. Note that the classes in the MathNet.Numerics.IO library are more appropriate for storing and loading data.

Statistics:
 More consistent behavior for empty and singleelement data sets: Min, Max, Mean, Variance, Standard Deviation etc. no longer throw exceptions if the data set is empty but instead return NaN. Variance and Standard Deviation will also return NaN if the set contains only a single entry. Population Variance and Population Standard Deviation will return 0 in this case.
 Reworked order statistics (Quantile, Quartile, Percentile, IQR, Fivenum, etc.), now much easier to use and supporting compatibility with all 9 Rtypes, Excel and Mathematica. The obsolete Percentile class now leverages the new order statistics, fixing a range check bug as side effect.
 New Hybrid Monte Carlo sampler for multivariate distributions. ~manyue
 New financial statistics: absolute risk and return measures. ~Phil Cleveland
 Explicit statistics for sorted arrays, unsorted arrays and sequences/streams. Faster algorithms on sorted data, also avoids multiple enumerations.
 Some statistics like Quantile or empirical inverse CDF can optionally return a parametric function when multiple evaluations are needed, like for plotting.

Linear Algebra:
 More reasonable ToString behavior for matrices and vectors:
ToString
methods no longer render the whole structure to a string for large data, among others because they used to wreak havoc in debugging and interactive scenarios like F# FSI. Instead, ToString now only renders an excerpt of the data, together with a line about dimension, type and in case of sparse data a sparseness indicator. The intention is to give a good idea about the data in a visually useful way. How much data is shown can be adjusted in the Control class. See also ToTypeString and ToVector/MatrixString.  Performance: reworked and tuned common parallelization. Some operations are up to 3 magnitudes faster in some extreme cases. Replaced copy loops with native routines. More algorithms are storageaware (and should thus perform better especially on sparse data). ~Thomas Ibel, Iain McDonald, Marcus Cuda
 Fixed range checks in the ThinQR decomposition. ~Marcus Cuda
 BUG: Fixed bug in Gram Schmidt for solving tall matrices. ~Marcus Cuda
 Vectors now implement the BCL IList interfaces (fixedlength) for better integration with existing .Net code. ~Scott Stephens
 Matrix/Vector parsing has been updated to be able to parse the new visual format as well (see ToMatrixString).
 DebuggerDisplay attributes for matrices and vectors.
 Map/IndexedMap combinators with storageaware and partially parallelized implementations for both dense and sparse data.
 Reworked Matrix/Vector construction from arrays, enumerables, indexed enumerables, nested enumerables or by providing an init function/lambda. Nonobsolete constructors now always use the raw data array directly without copying, while static functions always return a matrix/vector independent of the provided data source.
 F#: Improved extensions for matrix and vector construction: create, zeroCreate, randomCreate, init, ofArray2, ofRows/ofRowsList, ofColumns/ofColumnsList, ofSeqi/Listi (indexed). Storageaware for performance.
 F#: Updated map/mapi and other combinators to leverage core implementation, added nz variants where zerovalues may be skipped (relevant mostly for sparse matrices).
 F#: Idiomatic slice setters for submatrices and subvectors
 F#: More examples for matrix/vector creation and linear regression in the F# Samplepackage.
 More reasonable ToString behavior for matrices and vectors:
 Control: Simpler usage with new static ConfigureAuto and ConfigureSingleThread methods. Resolved misleading configuration logic and naming around disabling parallelization.
 Control: New settings for linear algebra ToString behavior.
 Fixed range check in the Xorshift pseudoRNG.
 Parallelization: Reworked our common logic to avoid expensive lambda calls in inner loops. Tunable.
 F#: Examples (and thus the NuGet Sample package) are now F# scripts prepared for experimenting interactively in FSI, instead of normal F# files. Tries to get the assembly references right for most users, both within the Math.NET Numerics solution and the NuGet package.
 Various minor improvements on consistency, performance, tests, xml docs, obsolete attributes, redundant code, argument checks, resources, cleanup, nuget, etc.
2.4.0  20130203
 Drops the dependency on the zlib library. We thus no longer have any dependencies on other packages. ~Marcus Cuda, Thomas Ibel
 Adds Modified Bessel & Struve special functions ~Wei Wu
 BUG: Fixes a bug in our iterative kurtosis statistics formula ~Artyom Baranovskiy
 Linear Algebra: Performance work, this time mostly around accessing matrix rows/columns as vectors. Opting out from targeted patching in our matrix and vector indexers to allow inlining.
 Linear Algebra: Fixes an issue around ThinQR solve ~Marcus Cuda
 Linear Algebra: Simplifications around using native linear algebra providers (see Math.NET Numerics With Native Linear Algebra)
 F#: Adds the BigRational module from the F# PowerPack, now to be maintained here instead. ~Gustavo Guerra
 F#: Better support for our Complex types (close to the F# PowerPack Complex type) ~Gustavo Guerra
2.3.0  20131125
 Portable Library: Adds support for WP8 (.Net 4.0 and higher, SL5, WP8 and .NET for Windows Store apps)
 Portable Library: New: portable build also for F# extensions (.Net 4.5, SL5 and .NET for Windows Store apps)
 Portable Library: NuGet: portable builds are now included in the main packages, no more need for special portable packages
 Linear Algebra: Continued major storage rework, in this release focusing on vectors (previous release was on matrices)
 Linear Algebra: Thin QR decomposition (in addition to existing full QR)
 Linear Algebra: Static CreateRandom for all dense matrix and vector types
 Linear Algebra: F#: slicing support for matrices and vectors
 Distributions: Consistent static Sample methods for all continuous and discrete distributions (was previously missing on a few)
 F#: better usability for random numbers and distributions.
 F# extensions are now using F# 3.0
 Updated Intel MKL references for our native linear algebra providers
 Various bug, performance and usability fixes