Making baby steps towards a finished thesis

Gotta love The PHD Comics, right?

I haven't yet finished my formal education: While I have been working at INDA for nearly 5 years I have also been studying Computer Science at University of Ljubljana. Officially I'm still a student!

With the project Sweden going on the question I hear a lot is: What about your degree? My intention was always to finish my studies and get a degree, but with the rigid bureaucracy of University of Ljubljana I can't graduate before the end of September, in spite my good intentions and time crunch. And that in combination with the fact that the lack of a degree isn't really hurting the job search process, means I'm in danger of dropping out!

Well I have decided not to let that happen! I here and now publicly announce my goal to graduate before 1st of October, 2013. With so few obligations left it shouldn't be that much of a challenge. If I make baby steps and do something about my thesis every week I should be there in no time!

I have two classes left, both of them due for completion at the end of this semester, but more importantly I still have to write the text part of my thesis. I have given my thesis some though and I have decided I'd love to do more research in one of my favourite fields of computer science: Soft Computing.

Warning: The next few paragraphs are more about computer science than they are about moving to Sweden. If that's not your cup of tea there is no shame in skipping them!

I have been fascinated by Genetics Algorithms since quite early in my academic career. Even though I didn't take a class on it since last academic year, my first attempts at solving problems with genetic algorithms date all the way back to my high school years.

I have to thank my friend and school mate Smily (who is doing some mighty impressive stuff of his own) for talking me into my very first introduction project, finding a linear equation for X using GA, that turned out to be heaps of fun. After that I used genetic algorithms for tackling the travelling salesman problem, and after that the project I'm submitting as my thesis.

A screenshot from my thesis project
NeuralCars

NeuralCars simulates a playground where agents (controlled by a artificial neural network) learn how to drive around a race track without human intervention. The process starts with a completely random weights in the feedforward neural network (10-8-8-2) but sure enough, given enough time selective breeding will create agents with impressive driving skills!

I've opensourced the project a while back, so you can check it out on Google Code. Compiled binaries are avaliable for download in case you'd like to compete against Skynet's getaway driver :)

So with the software part of the thesis nearly completed (my mentor did request a few changes) I just have to take the dive and start writing the text! Hopefully the pressure of public failure will keep my eye on the prize!

In other news I am being interviewed by Radio Sweden today! Thank you for creating enough buzz for the mainstream media to pick up on!