F#
F# is a functionally oriented programming language created by Don Syme and his team at Microsoft Research in Cambridge. Building on top of the .NET framework it augments the existing set of .NET languages (C#, C++, VB.NET, ...) with a functional counterpart. Although F# was inspired by OCaml it has evolved quite a bit and introduces a number of concepts unique to the language. In the summer of 2007, I did an internship with the F# team and have ever since been using the language for my machine learning research.
Math.Net Numerics
I have been actively contributing to a numerical library for the .Net platform since 2007. In 2009 our project dnAnalytics was merged with another open source library called Iridium by merging our feature sets to form the Math.Net Numerics open source numerical library for .Net.
Math.Net covers areas such as special functions, linear algebra, probability and statistics, random numbers, interpolation, integral transforms and more. The library targets .Net 3.5 and Mono, and in addition to a purely managed implementation will also supports native hardware optimizations, vectorization and parallelization. You can download Math.Net Numerics from Codeplex.
F# for Scientific Computing
F# in combination with the .NET runtime are a very strong platform for scientific computing. My current platform consists of F# in combination with visual studio (for compiling and interactive hacking), Math.Net and Intel MKL for fast linear algebra.
Here are some other resources and articles I find useful:
- Harness the Features of C# to Power Your Scientific Computing Projects: this is an interesting article about using the .NET platform for scientific computation. Two key points to remember are that although garbage collection incurs a bit of performance overhead, it could also improve runtime as it defragments the heap (better locality, smaller heap); also, your code becomes faster without doing anything as more and more JIT optimization is implemented.
- Optimize Managed Code For Multi-Core Machines: this article introduces the Microsoft Task Parallel Library, a very useful tool to leverage parallelism.