What's New in Version 4.0
- Full .NET 4.0 Support. Including samples and projects for Visual Studio 2010.
- Full F# 2.0 Support. Including more than 50 new samples.
- Multi-Core Ready. Many algorithms have been parallelized using the .NET Task Parallel Library.
- Many algorithms have been parallelized.
- New FunctionMath class that contains static extension methods for solving equations, optimization,
numerical integration and differentiation of functions encapsulated as delegates.
- Iterative algorithms have been streamlined.
- Vastly improved performance of linear program solver by several orders of magnitude.
- Implemented Parse/TryParse and decimal conversions on BigFloat.
- Performance improvements of up to 6x for BigInteger and BigFloat.
- We’ve added the most common window functions used in signal processing.
- New elementary functions: Clip (reduce to interval), fast IsNaN and IsFinite
(up to 50x faster on some architectures), Sqrt1pxm1 (accurate square root of
numbers close to 1).
- New special functions: Dawson’s integral, elliptic integrals, Fresnel functions, hypergeometric functions, Riemann zeta function.
Vector and Matrix Library
- The pure .NET linear algebra kernel has been parallelized.
- The Vector and Matrix classes contain methods in System.Math
and some elementary functions applied to vectors and matrices, respectively.
- Fill-reducing column ordering for sparse LU and Cholesky decompositions.
- Vectors and the rows or columns of matrices can be sorted.
- Min, Max, AbsoluteMin and AbsoluteMax methods have an overload that adds an
out parameter for the index of the extreme value.
- Improved performance of some managed level 2 and level 3 BLAS routines by up to 50%.
- New ComposedComplexVector and ComposedComplexMatrix classes to efficiently represent
vectors and matrices whose real and imaginary components are stored in real vectors and matrices.
- The computation phase of several statistical models has been parallelized.
- Continuous distributions have new members: Support and GetExpectationValue.