Visual Numerics today released version 4.0 of their native C# numerical analyis
library, the IMSL C# Numerical Library. I had a chance to chat with some of the
members of the product team last week, and to see some demos of the new version,
and here's what you can look forward to if you have need of some high-powered mathematical
help in your applications.
If you've got an engineering background, the IMSL name may sound familiar. That's
because this library is built on the same set of algorithms that has been around
for literally decades - it was first implemented in Fortran a long while ago (VNI
has been around for 35 years, making it ancient in terms of software companies).
That history gives you access to a tremendous range of functionality here, all implemented
in 100% native C# code. You get linear algebra, eigensystems, differential equations,
optimization, tons of statistics, random number generation, neural networks, financial
functions - the overview catalog is about 20 pages long. These are the things you'll
find necessary when you get into engineering and deep statistical applications, as well as the sort of financial modeling that drives the top forms on Wall Street.
The company isn't just resting on its laurels though. New features in version 4.0
include a new high-performance dense linear programming optimizer that beats some
established Fortran-based competitors, classification-oriented neural networks (the
sort of thing you need to do fast data mining and predictions in client code), and
an implementation of the Mersenne Twister random number generator algorithm. There's
also a 2D charting package coming later this year.
At the same time, VNI is also releasing version 4.0 of their pure Java JMSL Numerical
Library, letting you use the same tested algorithms in platform-independent portable
code. That's perhaps of less interest to my .NET-based readership but pretty darned
useful if you need to offload some things to client systems across a variety of
platforms.
I fondly remember IMSL from my own undergraduate days, and if I end up with some
knotty math to handle in my own .NET career I'll certainly turn here again. If this
sounds like the sort of thing that might be useful to you, you can
register to be contacted with an evaluation copy.