QuickStart Samples
Sample Applications
Get it now!
Download trial version
How to Buy
Contact Us
Our customers

This is a partial list of companies who are using our libraries:

ABB Robotics
Astra Schedule
Babson College
Canadian Council on Learning
Canyon Associates
Caxton Associates
Constellation Energy
Duke University
Engelhard Corporation
Equipoise Software
Galileo International
Global Matrix
The Hartford
Infinera Corporation
JDS Uniphase
LaBranche & Co.
Learning & Skills Council
Jacobs Consultancy
Litman Gregory
Lucas Systems
Malvern Instruments
Merck & Co.
Monitor Software
NanoString Technologies
Paletta Invent
Parametric Portfolio Associates
RATA Associates
Standard & Poor's
Strategic Analysis Corporation
Univ. of Alicante
Univ. of South Carolina
US Army

New Version 6.0!

Try it for free with our fully functional 60-day trial version.

Download now!

Extreme Optimization Numerical Libraries for .NET

What's New since Version 6.0

Version 6.0.18182.0 (July 2, 2018)

  • Upgrade native libraries to Intel® Math Kernel Library version 2018 Update 3.
  • ParallelOptions is now exposed for all algorithms to enable cancellation and other scenarios.
  • The NonlinearProgram class has a new constructor that accepts variable names.
  • Download version 6.0.18128.0Download this version.

Version 6.0.18128.0 (May 8, 2018)

  • Upgrade native libraries to Intel® Math Kernel Library version 2018 Update 2.
  • Binomial Generalized Linear Model can now be used with count data. NEW
  • Fix: Exception in nonlinear regression when error sum of squares is negative.
  • Fix: Exception in CSV reader.
  • Fix: Conversion to boolean should return true for nonzero values.

    Download version 6.0.18128.0Download this version.

Version 6.0.18065.0 (March 7, 2018)

  • Upgrade managed linear algebra library to LAPACK 3.7.0. NEW
  • Add R-style variations of quantiles. NEW
  • The threshold for parallel execution of vector maps can now be configured. NEW
  • Anderson Darling distribution is now public.
  • Fix: NoIntercept ignored in polynomial regression model.
  • Download version 6.0.18065.0Download this version.

Version 6.0.18046.0 (February 15, 2018)

  • Fix: Incorrect parameter transformation for power/exponential regression without intercept.
  • Download version 6.0.18046.0Download this version.

Version 6.0.18029.0 (January 29, 2018)

  • Workaround for 64 bit RyuJIT code generation bug in .NET 4.7.1.
  • Add name of native provider in hybrid provider description.
  • Download version 6.0.18029.0Download this version.

Version 6.0.17348.0 (December 14, 2017)

  • Fix: Index is lost when removing missing values from a vector.
  • Download version 6.0.17348.0Download this version.

Version 6.0.17249.0 (September 8, 2017)

  • Upgrade native libraries to Intel® Math Kernel Library version 2018 Update 0.
  • Fix: Time offset is wrong for weekly recurrences.
  • Fix: Student t inverse CDF is wrong for some inputs.
  • Fix: Race condition in parallel MIP solver.
  • Download version 6.0.17249.0Download this version.

Version 6.0.17150.0 (May 30, 2017)

  • Augmented Dickey-Fuller test. NEW
  • The Extreme.Data and Extreme.Data.Json assemblies are now also strongly named.
  • Improve performance for level 2 managed sparse BLAS.
  • Fix: Race condition in parallel MIP solver.
  • Download version 6.0.17150.0Download this version.

Version 6.0.17135.0 (May 15, 2017)

  • Upgrade native libraries to Intel® Math Kernel Library version 2017 Update 3.
  • Improve performance for various vector operations.
  • Fix: Some dense matrix multiplications are very slow.
  • Improved support for custom aggregators based on accumulators.
  • Download version 6.0.17135.0Download this version.

Version 6.0.17121.0 (May 1, 2017)

  • Compute factors of symmetric and Hermitian indefinite decomposition. NEW
  • Performance improvements to Linear Program solver.
  • Fix: Round-off error causes exception in SimpleRegressionModel.
  • Download version 6.0.17121.0Download this version.

Version 6.0.17087.0 (March 28, 2017)

  • Support for Visual Studio 2017. NEW
  • Fix: Exception when solving mixed integer linear programs.
  • Download version 6.0.17087.0Download this version.

Version 6.0.17058.0 (February 27, 2017)

  • Jacobi elliptic functions. NEW
  • The generic Operations<T> class has been optimized to eliminate nearly all overhead for the most frequently used operations on the most common argument types.
  • Fix: Weighted aggregators return NaN when first weight is 0.
  • Fix: Column names are not set when setting the column index of a data frame.
  • Fix: Potential stack overflow in OneSampleTest.Summarize().
  • Download version 6.0.17058.0Download this version.

Version 6.0.17031.0 (January 31, 2017)

  • Exponentiation operator for all numerical types in Visual Basic .NET. NEW
  • ConditionalVariances property on GARCH models. NEW
  • Fix: Exception when fitting higher order GARCH models.
  • Fix: Maximum likelihood for GARCH models not set correctly.
  • Download version 6.0.17031.0Download this version.

Version 6.0.16354.0 (December 19, 2016)

Version 6.0.16343.0 (December 8, 2016)

Version 6.0.16328.0 (November 23, 2016)

Version 6.0.16322.0 (November 17, 2016)

Version 6.0.16312.0 (November 7, 2016)

  • Broadcasting vectors in matrix operations. NEW
  • Upgrade native libraries to Intel Performance Libraries version 2017 Update 1.
  • Fix regression in nonlinear program solver.
  • Download version 6.0.16312.0Download this version.

Version 6.0.16305.0 (October 31, 2016)

  • Cramer-von Mises Goodness-of-fit test. NEW
  • Polygamma function. NEW
  • Modified Bessel functions of real order. NEW
  • Symbolic constraints that are linear in the variables are now recognized as such.
  • The Nonlinear Program solver can now recover when it encounters an infeasible subproblem.
  • Download version 6.0.16305.0Download this version.

Version 6.0.16298.0 (October 24, 2016)

  • Improve range and accuracy of matrix exponential.
  • "Partial application" functions for incomplete and regularized Gamma and Beta functions. NEW
  • Vector Map methods that include index as delegate argument. NEW
  • Fix: Quadratic program solver does not always honor variable bounds.
  • Fix: Higher degree polynomial fit throws exception.
  • Fix: Some weighted simple regression results are wrong.
  • Download version 6.0.16298.0Download this version.

Version 6.0.16290.0 (October 16, 2016)

  • Zernike polynomials. NEW
  • New overloads for numerical integration methods that take Interval objects to specify bounds.
  • Fix: Covariance matrix for Levenberg-Marquardt optimizer is scrambled.
  • Fix: Simple regression model with 0 degrees of freedom throws exception.
  • Fix: Polynomial regression with vector input doesn't give full polynomial result.
  • Download version 6.0.16290.0Download this version.

Version 6.0.16283.0 (October 9, 2016)

Version 6.0.16255.0 (September 12, 2016)

  • Upgrade native libraries to Intel Performance Libraries version 2017.0.
  • Enable Conditional Numerical Reproducibility option for native libraries.
  • NEW
  • Fix: Regression in linear program pre-solve.
  • Download version 6.0.16255.0Download this version.

Version 6.0.16241.0 (August 29, 2016)

  • LeastSquaresOptimizer base class for nonlinear least squares algorithms. NEW
  • Trust Region Reflexive algorithm for nonlinear least squares. NEW
  • Trust Region Reflexive algorithm option in nonlinear curve fitting. NEW
  • Improved documentation for nonlinear least squares algorithms.
  • The Cholesky decomposition of diagonal matrices has been optimized.
  • The performance of ARIMA model fitting has been improved.
  • Missing matrix methods ReduceRowsInto and ReduceColumnsInto have been added.
  • Fix: An issue with permutations of a symmetric matrix has been fixed.
  • Download version 6.0.16241.0Download this version.

Version 6.0.16231.0 (August 18, 2016)

  • Bug fix release.
  • Download version 6.0.16231.0Download this version.

Version 6.0.16195.0 (July 14, 2016)

  • The sort order of the category index can now be specified when creating categorical variables. NEW
  • New methods to get a vector of samples from a discrete distribution. NEW
  • Inverse hyperbolic functions for decimal and quad precision numbers. NEW
  • The accuracy of inverse hyperbolic functions has been improved.
  • Some performance improvements when constructing data frames.
  • Methods for computing Savitsky-Golay coefficients have been moved to the Smoothing class.
  • Fix: Modified model input is now correctly handled when recomputing a model.
  • Download version 6.0.16195.0Download this version.

Version 6.0.16184.0 (July 3, 2016)

  • Limited Memory BFGS Optimizer. NEW
  • Fix: Ridge parameter has no effect.
  • Fix: Loss of accuracy in inverse regularized beta function.
  • Fix: Index of sorted inner joins is blank.
  • Download version 6.0.16184.0Download this version.

Version 6.0.16170.0 (June 20, 2016)

  • Data Access Library
  • Unified API for reading and writing data frames, matrices, and vectors. NEW
  • Reading and writing R's .rda/.rdata and .rds files. NEW
  • JSON serialization using JSON.NET. NEW
  • Other supported formats include: delimited text (CSV, TSV...), fixed-width text, Matrix Market, Matlab®, stata®
  • Download version 6.0.16170.0Download this version.

Version 6.0.16157.0 (June 7, 2016)

  • Two-way ANOVA: support for Type I, Type II, and Type III sums of squares NEW
  • Fix: Linear Program pre-solve doesn't respects all tolerances.
  • Fix: Some improvements to handling vectors and matrices with 0 elements.
  • Download version 6.0.16157.0Download this version.

Version 6.0.16137.0 (May 18, 2016)

Version 6.0.16120.0 (May 1, 2016)

Version 6.0.16113.0 (April 25, 2016)

Version 6.0.16106.0 (April 18, 2016)

Version 6.0.16099.0 (April 11, 2016)

  • Use R-style model formulas to specify statistical models. NEW
  • QuickStart samples updated to show using model formulas.
  • Automatically infer if logistic regression is binomial or multinomial.
  • Option to specify encoding when converting to a categorical vector.
  • Download version 6.0.16099.0Download this version.

Version 6.0.16096.0 (April 8, 2016)

  • All Vector.Cast<> conversions between numerical types now work as expected.
  • Add Vector.Cast<> conversion to boolean and to string.
  • DataFrame.FromObjects option to specify field for row index. NEW
  • Fix: some translation errors in auto-translation of some LAPACK routines.
  • Download version 6.0.16096.0Download this version.