Parallel Machine Learning for Hadoop/Mapreduce – A Python Example

Posted: February 8th, 2010 | Author: Soren Macbeth | Comments

source

Wild Thing

Posted: February 5th, 2010 | Author: Soren Macbeth | Comments

sorenmacbeth posted a photo:

Posted via email from Soren Macbeth’s posterous

(fab) – a pure javascript DSL for building async web apps

Posted: February 5th, 2010 | Author: Soren Macbeth | Comments

source

Measuring Measures: Learning About Statistical Learning

Posted: January 15th, 2010 | Author: Soren Macbeth | Comments

source

The APInternet

Posted: January 13th, 2010 | Author: Soren Macbeth | Comments

I love APIs. I love that pretty much any useful web application has an API. I love hacking around mashing up APIs building little prototypes for things in a few hour. This is actually how StockTwits was first born. I hacked together something using the Twitter API and the Tumblr API (A great example of an MVP before I had ever heard about that term). Doing this involves finding a service that has cool data, finding it’s API docs and/or searching for someone else that has already written a ruby or python wrapper and then playing around getting the data I want.

I would really like for this process to be more automated. There should be a metatag that points to a machine readable description of a services API, similar to how RSS feeds can be automatically discovered now. Crawlers could then be built to discover and harvest information from APIs, etc etc. The possibilities are endless from there.

Obviously this is a hard problem to solve. APIs vary greatly, write a machine readable description language would be a pain in the ass. That is why I don’t want to do it. However, I bet there is someone out there that is up to the challenge.