Last week on Wednesday I kickstarted the first AI coding dojo at our lab with a group of interested friends. The goal is to organize a group of interested students with different backgrounds and train to implement Artificial Intelligence algorithms. In a coding dojo the goal is not to build working software but to improve our coding and style. Why especially on Artificial Intelligence? Well in my opinion implementing AI algorithms is much different from implementing regular software. While doing so we also try to improve our coding and style and implement everything in a test driven manner. The setup is as follows. One person brings an interesting algorithm that is the topic of the dojo. It should not be too hard since we restricted a session to 2 hours. In the first session we implemented Expectation Maximization for a Mixture of Gaussians. The participants of the dojo gather around a projector or large screen connected to a single computer. Two participants sit on the computer and implement part of the problem using pair programming. The rest of the group watches what happens on the screen and gives helpful comments when needed. The two programmers switch after 10 minutes.
The first dojo went slow but I think people learned from each other and enjoyed it. The next dojos will include advanced search algorithms (D*), collaborative filtering (Matrix Factorization) and the cognitive architecture SOAR.
The status of this week is in: https://github.com/dkohlsdorf/Artificial-Intelligence-Dojo
No comments:
Post a Comment
Note: Only a member of this blog may post a comment.