Probabilistic Robotics

It’s been a long haul, but I’m done with Probabilistic Robotics, the beginning of my reading spree until the end of my time here at Columbia. If you don’t remember, I decided I should start a reading spree until I didn’t have access to the huge library at Columbia: hell, all that money has to be paying for something. But the fact that I said that back in October, and only now finished the first book says something about… something. It’s kind of complicated.

When I said I’m done with the book, I didn’t mean that I finished it from cover to cover. No, by the time I was halfway through the book it stopped being fun and started being tedious. The content of the book is relevant to my interests, but it was written in such a way that it didn’t really make sense how to transform the concepts to code. The psuedo-code was very math heavy, whereas lots of single letter variables represented potentially huge covariance matricies. My inexperience with matrix math didn’t help here: matrices were used everywhere, and it wasn’t immediately apparent how it all fit together.

I suppose the fact I didn’t try implementing any of this in code didn’t help. I know, big mistake on my part, but it seemed kind of… pointless. I don’t have any motivating factors right now, and the only thing I would really be interested in implementing would be the FastSLAM algorithm, but I couldn’t understand anything in that chapter (yes, I’ll take that hit to my geek standings. I did not understand something in a book).

Another topic of contention was my experience with probability: the author assumed a bit more familiarity with probability than I actually had, so I was kind of lost for much of the time. Sure, I know what the covariance is, but keeping everything straight in addition to the covariance wasn’t easy.

Hence, my next book will be a book on Bayescraft, and getting myself familiar with probabilistic ideas. Since I’m in the middle of a Devfest hackathon, I can’t exactly get up and go get a book, but I’ll be sure to have it waiting for me when the storm ends.

And just FYI, sometime, I’m going to read Probabilistic Robotics again and win.

Also, I should note I have no kindle. It’s okay, it means I’m not trying to hack it into my life, which would undoubtedly sink huge amounts of time.

EDIT: Right! I forgot to talk about Devfest! The Devfest at which I played a non-trivial role! I didn’t botch my role as workshop facilitator, which is a big win, although I need to learn how to wield a big stick and get people to stop talking, either on stage or off it. Aside that, the tutorials covered lots of ground in relatively little time (it wasn’t necessarily good: some of the audience’s eyes were glazing over, although I should note that most of the audience was a very quiet crowd. It gives me a new appreciation for a responsive audience). So yeah, cool. My big contribution to Devfest is made, and from here on out I essentially just have to code. So far, I just have a gimp plugin published, so… I guess I’ll get started on another one. Good times.