Thursday, July 26, 2007

OSCON: Machine learning with Perl

The last talk of the day is Machine Learning Made Easy with Perl with Lino Ramirez. When I noticed this one on the programme I immediately put it in the "must go to" pile as, of course, for my day job I build autonomous agent based systems to manage intelligent robotic telescope networks.

Lino says,

It's all about empowering people

You can think about machine learning as a three phrase process: preparation (definition, gathering, analysis, cleaning, selection), modelling (selection, development, evaluation) and implementation (evaluation, implementation). After implementation there should always be a way for the system to keep learning.

Update: He's currently going through an example using stock quotes, talking about clustering of data. I'm having flashbacks to yesterday's data mining tutorial. If you're interested in following along, apparently the code will be available for download after the talk.

Update: We've been pointed at "Knowledge-Based Clustering" by Witold Pedrycz as a good text on clustering algorithms. I think I've actually got this on my shelf at home, but I'm not sure I've ever done more than flip through it, I'll have to dig it out when I get back into the UK.

Update: Apparently he's a PDL and PGPLOT user. I'm feeling right at home here, this guy is definitely a an academic...

Update: His next (and final) example is based on medical diagnosis, and he's talking about support vector machines classifiers and his PDL implementation of the technique. I'm going to have to mail him at the end of this and get the source code to this, or persuade him to tidy it up and release it onto CPAN.

Update: If you want to code and the slides from this talk you need to email the speaker at ramirez@aranducorp.com.

3 comments:

  1. Given that this is an Open Source conference, do you know whether O'Reilly plan to release the slides for ALL the talks? I can find no mention of it on the website.

    ReplyDelete
  2. Generally its down to the individual authors, a lot of the talks never show up. If they do get released, they end up on the conference website.

    ReplyDelete
  3. Jon: as al says, we ask the speakers to upload their slides. Because many people present from their laptops, the slides don't all pass through us. We try, and those we get them from we put online. If there's a speaker who didn't put their slides online but whose slides you really wanted to see, send me mail (gnat AT oreilly DOT com) and I'll bug them for you.

    ReplyDelete