Rewriting a Logistic Regression from C# to F#, part 2

In our previous post, we looked at James McCaffrey’s code, “Gradient Descent Training Using C#” from MSDN magazine, and took a stab at rewriting the first part in F#, to clarify a bit the way the dataset was created. Today, we’ll dive in the second block, which implements the logistic regression using gradient descent. Again, we won’t discuss why the algorithm works – the article does a pretty good job at that – and focus instead purely on the F# / C# conversion part.

Rewriting a Logistic Regression from C# to F#, part 1

I will admit it, I got a bit upset by James McCaffrey’s column in MSDN magazine this month, “Gradient Descent Training Using C#”. While the algorithm explanations are quite good, I was disappointed by the C# sample code, and kept thinking to myself “why oh why isn’t this written in F#”. This is by no means intended as a criticism of C#; it’s a great language, but some problems are just better suited for different languages, and in this case, I couldn’t fathom why F# wasn’t used.

First impressions with DiffSharp, an F# autodiff library

A few weeks ago, I came across DiffSharp, an automatic differentiation library in F#. As someone whose calculus skills have always been rather mediocre (thanks Wolfram Alpha!), but who needs to deal with gradients and the like on a regular basis because they are quite useful in machine learning and numerical methods, the project looked pretty interesting: who wouldn’t want exact and efficient calculations of derivatives? So I figured I would take a couple of hours to experiment with the library. This post is by no means an in-depth evaluation, but rather intended as “notes from the road” from someone entirely new to DiffSharp.

Fun with L-Systems

I had the great pleasure to speak at CodeMash this week, and, on my way back, ended up spending a couple of hours at the Atlanta airport waiting for my connecting flight back to the warmer climate of San Francisco – a perfect opportunity for some light-hearted coding fun. A couple of days earlier, I came across this really nice tweet, rendering the results of an L-system:

{start:'FFPF',rules:{F:'PF++F[FF-F+PF+FPP][F]FFPF',P:''},'α':60} pic.twitter.com/JZGDV4ghFy

— LSystemBot 2.0 (@LSystemBot) January 10, 2015

The 2014 F# Tour in numbers

Well, we are in the last hours of 2014, and I am nearly recovered from the craziness that was the F# Europa Tour 2014, so here we go – the Tour, in cold, hard facts (after all, I am a numbers’ guy):

• 40 days of travelling across Europe.
• 16 talks.
• 5 workshops (about 50 hours total).
• 9 countries.
• 6991 miles (11,250 kilometers) travelled, roughly (this is straight-line city to city, so the actual number is probably a good deal larger).
• 14 hours of bus.
• roughly 50 hours of train.
• roughly 28 hours of plane.
• 12 cities visited (and spoken at!).
• I lost track of how many gallons of beer were ingested. This is big data.
• 500 attendees? Maybe more? See previous data point.
• Delivered hundreds of shiny fsharp.org stickers to F# Communities across Europe. Btw, in case you didn’t hear - the F# Software Foundation is now a full-fledged, legally established entity, and YOU can be a member. Check it out!