At work, I’ve been working on motion detecting algorithms for video games, which involves a fair amount of statistics and machine learning. So I’ve also been learning a lot more about statistics, data mining, dealing with noisy data (we have some data that’s reliably tags, and a lot of data whose tags are likely to be wrong), comparing different methods, explaining these things to non-technical people, visualizing the data (distributions, scatterplots), etc. I’m still feeling like an idiot a lot of the time, but the system (most of it implemented in Python) works quite well, I just keep having the nagging doubt that there’s some algorithm out there that give better result, some analysis method that will allow me to see more in the data, etc. But with deadlines looming, I can’t afford too much analysis paralysis either.
I’ve also been asking questions on Cross Validated, and browing back through the answers of one of my old questions, I saw the best one was answered by a certain “John Salvatier”, and was thinking, “hmm, that name sounds familiar …”
On the side (I’m another one of those “program by day, and program some more by night” guys), I’m working on a little web-based game that’s not in a state worth showing right now, but is going along smoothly so far.
At work, I’ve been working on motion detecting algorithms for video games, which involves a fair amount of statistics and machine learning. So I’ve also been learning a lot more about statistics, data mining, dealing with noisy data (we have some data that’s reliably tags, and a lot of data whose tags are likely to be wrong), comparing different methods, explaining these things to non-technical people, visualizing the data (distributions, scatterplots), etc. I’m still feeling like an idiot a lot of the time, but the system (most of it implemented in Python) works quite well, I just keep having the nagging doubt that there’s some algorithm out there that give better result, some analysis method that will allow me to see more in the data, etc. But with deadlines looming, I can’t afford too much analysis paralysis either.
I’ve also been asking questions on Cross Validated, and browing back through the answers of one of my old questions, I saw the best one was answered by a certain “John Salvatier”, and was thinking, “hmm, that name sounds familiar …”
On the side (I’m another one of those “program by day, and program some more by night” guys), I’m working on a little web-based game that’s not in a state worth showing right now, but is going along smoothly so far.