1. To learn through failure. I flearnt how to ski by sliding face first down a mountain.
  2. A machine learning library written in F# for F#. Woah, FLearn learned that pattern quickly!

Built on top of the Math.Net Numerics linear algebra library, FLearn offers an extensible, scalable and interactive take on machine learning. Utilizing tools from online convex optimization, FLearn provides a framework for specifying and optimizing various predictive models. Straight out of the box you can:

  1. Construct powerful predictive models, with statistical and worst case guarantees, via the minimization of convex loss functions.
  2. Utilize powerful feature maps such as those provided by random fourier features and the nystrom method.
  3. Make predictions and estimate their error! Think error bars and gaussian processes....
  4. Learn from noisily labeled data via corruption corrected loss functions.

The source code for FLearn is available here.


Fake it till you make it: Haskell like Type Classes for F# together with some basic language extensions. I make no claims to its originality and have shared it here purely for convenience. If I have unjustly stolen your ideas please get in touch.

The source code for FUnhinged is available here.