What's New
Below is a nonexhaustive list of the major features that were added in recent upgrades. Smaller improvements and additions are not in this list.
New in version 4.2 (December, 2011)
 Automatic differentiation
 Supply functions and constraints as lambda expressions and have gradients computed automatically.
 Backward differentiation with common subexpression elimination for optimal performance.

Extensible with builtin support for methods of System.Math and most elementary and special functions in the library.
 Orthogonal polynomials Chebyshev, Gegenbauer, Hermite, Laguerre, Legendre polynomials and sequences.
 Stepwise regression Automatic variable selection for linear regression.
 Contingency tables Both 2x2 and RxC tables are supported.
 Improved SQL Server support Easier deployment.
New in version 4.1 (July, 2011)
 Optimization framework. Provides a generic model for defining and solving optimization problems.
 Quadratic Programming. Solve quadratic optimization models with linear constraints.
 Nonlinear Programming. Optimize nonlinear functions with linear or nonlinear constraints.
 New Decimal functions extend all the functions in System.Math to the decimal type, including sin, cos, exp.
 Improved elementary functions. Evaluate sine, cosine and tangent accurately for huge arguments.
 Iterative sparse solvers Efficiently solve systems with many thousands of variables, optionally using preconditioners.
 New probability distributions LogSeries and Maxwell.
New in version 4.0 (November, 2010)
 Full .NET 4.0 Support. Including samples and projects for Visual Studio 2010.
 Full F# 2.0 Support. Including more than 50 new samples.
 Multicore Ready. Many algorithms have been parallelized using the .NET Task Parallel Library.
 Improved sparse solvers using fillreducing column orderings.
 New Sparse Linear Program Solver can solve problems with more than 1 million variables.
 Mixed integer programming.
 New Special functions. Hypergeometric, elliptic integrals, Fresnel, Riemann zeta.
 FFT Window Functions.
New in version 3.6 (February, 2010)
 Singlepreicision vector and matrix library
 Nonnegative Matrix Factorization (NMF)
 Nonnegative Least Squares (NNLS)
 Numerical integration in 3 or more dimensions
New in version 3.5 (October 2009)
 Full .NET 3.5 Support. Including samples and projects for Visual Studio 2008.
 Ordinary Differential Equations. Integrate stiff and nonstiff systems of Ordinary Differential Equations (ODE's) using our stateofthe art algorithms.
 Improved Curve Fitting. We made our algorithms more robust, and added new features, including confidence and prediction bounds.
 Time Series Models. Exponential smoothing and ARIMA (AutoRegressive Integrated Moving Average) models.
New in version 3.1:
 Arbitrary precision numbers. BigInteger, BigRational, BigFloat.
 Generic Arithmetic Framework. Write algorithms that can use any numeric type.
 Generic Linear Algebra. A complete generic vector and matrix library.
 2D Numerical Integration.
 Generalized Linear Models. Poisson regression, probit regression and other variants.
New in version 3.0:
 **2D Fast Fourier Transforms. **Compute 2dimensional FFT’s using managed or native code.
 Multivariate Statistical Analysis. Principal Component Analysis (PCA), Kmeans Clustering, Hierarchical Clustering.
 Multivariate Probability Distributions. Multivariate normal and Dirichlet distributions.
New in version 2.1:
 Sparse Matrix Library. Efficiently calculate with huge, sparse matrices.
 Linear Programming. Our dense LP solver is second to none.
 Fast Fourier Transforms. Compute 1dimensional FFT’s using managed or native code.
New in version 2.0:
 Matrix Debugger Visualizer. Inspect the elements of a matrix at debug time in table form. (screen shot)
 Generic interfaces. For example, all collection classes support the appropriate IList<T> interface.
 New structured matrix types. Perform calculations on band matrices and diagonal matrices more efficiently.
 Sort and filter data. New methods give you complete control of which observations are included in your statistical calculations.
 Logistic Regression. Predict binary outcomes in terms of one or more variables.
 Nonlinear Regression. An extension of our nonlinear curve fitting classes that gives you full access to the statistical properties of your model.