Phillip Trelford's Array

POKE 36879,255

NorDevCon 2014

NorDevCon is a one day agile and tech conference held annually in the historic English town of Norwich, also the setting for the recent British comedy Alan Partridge: Alpha Papa.

Last Friday I took the short hop across the Fens to Norwich to talk about Data Science and Machine Learning with F#. First a talk on F# Type Providers entitled All your base types are belong to us (thanks to Ross McKinlay for the meme), and then after lunch a hands on Machine Learning workshop exploring `Titanic passengers using data from Kaggle (a data science company based in San Francisco).

The conference attracted a great range of speakers with some really interesting sessions:

NorDevCont

In the morning I got to chat with Jon Skeet about programming with kids, and then watched him answer stack overflow questions at a frightening pace.

Jason Gorman gave a lively and warm opening keynote on Software Apprenticeships, which included a Skype session with his brave apprentice, Will Price. Immediately followed by Chris O’Dell on Continuous Delivery at 7Digital which ended with a lot of interested questions.

In the afternoon I caught Phil Nash’s thought provoking session on Agile and Mobile – do they work together as well as they should. Phil talked about a schism in testing approaches on mobile platforms, particularly iOS, with some in the community advocating unit testing and TDD and others none at all. Check out Phil’s links from the talk to learn more.

The closing keynote came from the highly respected Nat Pryce and Steve Freeman on Building SOLID Foundations. The talk focused on design principles for addressing complexity in mid-scale codebases. Nat gave examples of successfully taming complexity in an unnamed risk management project using immutability and DSLs, in effect a functional-style approach. An approach that Jessica Kerr also explored in some depth in her popular Functional Principles for Object-Oriented Developers talk at last year’s Øredev.

The day was capped off was a hearty dinner with a fun format where people moved around between the courses.

All your types are belong to us!

My first talk show cased typed access to a vast array of data sources through to software environments like Hadoop and R, via F# Type Providers:

All your types are belong to us! from Phillip Trelford

Type Providers covered:

  • JSON, CSV, HTML Tables and the World Bank (via FSharp.Data)
  • SQLite (via Ross’s SQL Provider which also supports MySQL, Oracle, Postgre & SQL Server)
  • R (via BlueMountain Capital’s R-Type Provider)
  • Hadoop (in the browser via Try FSharp)

All of the Type Providers shown are open source projects developed by the F# community, can be easily integrated into projects via Nuget and run on Linux, Mac and Windows.

The HTML Table type provider being the most recent, developed by Colin Bull, it gives immediate typed access to data in tables on web pages.

After the talk Jon Skeet suggested a Protocol Buffers Type Provider:


For which it turns out Cameron Taggart already has a project called Froto.

If you are interested in creating your own Type Provider I’d recommend reading Michael Newtons’s Type Provider tutorial. Michael will be running a free workshop on building Type Providers at the F#unctional Londoners meetup on May 1st.

Hands On Machine Learning workshop

This session gave an introduction to machine learning, using F#’s REPL and the CSV Type Provider to easily explore data on Titanic and predict survival outcomes:

Machine learning from disaster from Phillip Trelford

It was a great group and everyone managed to complete the task and produce good prediction results using decision tree learning in just 1.5 hours.

Wrapping Up

Thanks to Paul Grenyer for organizing an excellent conference and giving me the opportunity to speak. Paul put together a really interesting programme which was extremely well-executed.

A++ would recommend :)

Pingbacks and trackbacks (1)+

Comments are closed