Numerical computing with F# and MKL - Adam Mlocek

.NET numerical libraries tend to be designed in a heavily object oriented style. For someone used to the expressiveness of numpy, Matlab or R this might be a bit of a problem. Also, even if some functionality is delegated to high performance libraries like Intel Math Kernel Library, the design does not allow us to evaluate vector/matrix expressions in the most efficient way or use all available memory.

In this talk I would like to present an open source F# math and stat library which tries to address all those issues. We will see how F# overloaded operators and functions, strong and static typing and lazy expression evaluation can give us the perfect mix of intuitive API and high performance, all in your favourite language!

If time permits I will show the recent additions to the library: strongly typed statistical data frame and F# implementation of Generalised Linear Model and compare its performance to Revolution R and H2O.

The library is available here: https://github.com/Statfactory/FCor (https://github.com/Statfactory/FCore)