Machine Learning with F# | Hacker News Reader

Machine Learning with F#

fsharp.org

84 pointssytelus18 comments

This blog post linked from the OP is very nice:

http://clear-lines.com/blog/post/Support-Vector-Machine-in-F...

F# looks lovely, but I have yet to find good cross-platform 2D plotting and 3D visualization story.

(ActiViz would give me everything I need, but the current wrappers are several releases out of date - though there is an ongoing Kickstarter to hopefully fix that)

F# has been successfully used by some of the most advanced machine learning teams in the world, including the Microsoft Research Machine Learning and Applied Games groups at Microsoft Research

These may be some top-notch groups... but, it's a red flag (to me) that they didn't include anyone without a vested interest in the language as part of this list. This just says that the company that developed the language also uses it.

Its use is increasing among .NET shops in the financial sector, the same area where Haskell and OCaml get used.

Me, I use it mostly as a better scripting language.

outside of Jane Street (http://www.janestreet.com/technology/ocaml.php), I've not heard of functional languages being used in the financial sector. Where else is it used?

In Skills That Matter presentations they sometimes refer to deployments in the city (London finance sector).

You also have some use cases here,

http://msdn.microsoft.com/en-us/vstudio/gg634701#Who_Uses_F_...

http://fsharp.org/testimonials/

It's "Skills Matter" -- they're a company that offers training events, and they handle many of the big F# training events in NYC and London:

http://skillsmatter.com/

Oh right, thanks for the correction.

Most quants use functional programming to some degree when not locked down into C++, be it in a specifically funcational language or not. At least in London F# is commonly used in Credit Suisse, Barclays and UBS that I know of. Heard that Morgan Stanley uses it a bit too but didn't sound very widespread there.

Standard chartered

Credit Suisse

Barclays capital

Tsuru capital

GSA capital

Alphaheavy

Edf energy trading

... Just off the top of my head :-)

Barclays, HSBC and Credit Suisse are three place I know of just based on talking people who've worked there.

Novus in NYC uses scala and open sources some libraries (salat)

Do you know whether numerical analysis has improved with .NET? I know Morgan Stanley was evaluating it a few (4? yow) years ago, but there were two main problems:

- There were very few numerical libraries, and those that existed either had a high overhead (compared to, say, calling NAG from C++), or weren't very robust (didn't degrade well in edge cases).

- A windows license for a server machine with > 32 GB of RAM cost an arm and a leg, so server farms got very expensive very quickly. (Oddly, the price of the license depended on the RAM of the machine involved.)

It looks like NAG now has a .NET version, so hopefully things have improved. For a while, it seemed like math in .NET was strictly amateur hour, but I'm sure it's better now. I don't have the strength of will to research if the licensing cost has improved at all.

"A windows license for a server machine with > 32 GB of RAM cost an arm and a leg, so server farms got very expensive very quickly. (Oddly, the price of the license depended on the RAM of the machine involved.)"

FWIW, you now have these options in Windows Azure:

- Memory Intensive VM (4 x 1.6GHz CPU, 28GB RAM, 1,000GB Storage) : $0.90/hr

- Memory Intensive VM (8 x 1.6GHz CPU, 56GB RAM, 2,040GB Storage : $1.80/hr

http://www.windowsazure.com/en-us/pricing/calculator/?scenar...

Interesting. I've heard good things about the Windows Azure program, but I don't think any cloud-based option will work for finance. People generally want to keep their servers and data in-house, for trade secret protection and regulatory purposes.

On the type of projects I work for, license costs are rounding errors.

When you do big enterprise projects, and finance is such a case, there many higher costs associated with the project than what is paid in licenses.

IMO, dogfooding[1] is a good thing and is important to provide early feedback on new features before shipping them to customers (or releasing a beta).

F# has been used within Microsoft to implement TrueSkill[2] (the ranking system for Xbox Live) as well as the advertising platform for Bing[3].

[1] https://en.wikipedia.org/wiki/Eating_your_own_dog_food

[2] https://blogs.msdn.com/b/dsyme/archive/2012/04/19/updated-ve...

[3] https://www.simple-talk.com/opinion/geek-of-the-week/don-sym...

Well, presumably they're more interested in efficiently implementing their algorithms than in using other MSR technologies for their own sake, but you've got a point. See http://fsharp.org/testimonials/ for lots of other companies using F#. "Machine learning" only is only mentioned a couple of times, but it sounds like plenty of other companies are doing related work with F# (e.g. Kaggle mentions using it for "data analysis").

(disclosure: I'm doing some contract work with the F# team)

Nice. http://fdatamining.blogspot.com (linked under "Other links") is another good place for F# machine learning.

Reply on news.ycombinator.com