Day #252: Dead…line
Today is the day, my group presentation is due. Is it done? Well… sort of? Okay not quite, but we’re getting there. My group member still sucks, but since I yelled at him (in a professional manner of course), he’s gotten a little better and has been more responsive, so what’s the hold up? What a great question, let me explain.
Like the good time wizard that I’m turning into, I finished my homework 4 assignment yesterday and already submitted it. It helped having a clear goal for what I was planning on doing for the design your own problem portion of the assignment (which has been the most enjoyable parts of my homeworks for this class). However, we still don’t have everything ready for the presentation today.
Processing data takes time, when we train a kalman filter we have to iterate between making a prediction, updating the prediction, doing expectation maximization, then repeating this process all over again. It takes roughly 5 minutes for one iteration to finish and each time we do this process we iterate around 100 times. It depends on what we set our convergence criteria to, but we’ll get into that some other time when I’m not scrambling to get stuff done for school.
Point being, things take time. While he was training one Kalman filter, I was doing the other. It takes mine about 3 hours or so to converge and since he’s doing a different step he can get that part done in roughly 2 hours. We had 6 subjects we needed to process and right now we have 3 done completely and one that should be finished soon. The 6th one we haven’t started yet and will definitely not finish today, but the 4th and 5th one we’ll get done. It will be really close, but we will get it done. Four should be done here shortly, number 5 will be done … gulp… about 10 minutes before class if we’re lucky.
So yeah! Fun times. Being the overachiever I am I saw a small next step that could be taken to improve our results and I went ahead and took it. The outcome was very impressive and while I think it can be improved further, so I’m very happy to have done it. The end result is an improved prediction without any added computational work so not only is this a good thing, but it won’t make this long drawn out process harder. I know I’m being cryptic about all this, but I can’t share it until after we publish, I’m sorry!
That said, it’s back to work I go. It’s going to be to the wire today and I need to get the slides sent off to the instructor prior to the class. While I may have time to introduce expectation maximization to everyone tomorrow, I think I’m going to opt instead to give an update about how today went. Oh and our next and final deadline for this class is Sunday, so I have a lot of work to do. Plus my biweekly review paper is due today (joy) so I’ll have to find time to do that in between all this other work. Expect to see that either tomorrow or the next day as well! How is there so much work when we have a pandemic going on?!?!