I didn't ever believe I could get so pointlessly frustrated with the need of probability distribution to sum to one... (even while knowing that it's actually helping me find logical flaws)

Aug 2: Weekend!!!

Bleagh... I'm very, very glad this week is over.

Today I got up and ate dorm breakfast of harusame and beef with something I can't remember haha (and miso soup and rice).  I got out of the house and to work on time, and worked a bit on my recommender business.  Unfortunately, the results of the recommender are not as I would expect, and it is far slower than regular PLSA to boot (it's supposed to be faster).  The big problem is that I now have a sheet of paper full of things that I am suspicious about, but Kabutoya-san wasn't there again today, and apparently noone else in the group works on topic models. D:  So, on Friday, I hope he is ready to answer a hundred million questions about what might be going wrong with the recommender.  It's a bit hard to be productive when you not only have serious doubts about the theoretical completeness of the model, but also about the suitability of your dataset to the problem. >_<

At lunch I ate murugi curry again (I have no idea who makes the schedule), because it was there, and looked better than everything else (except for pricey katsu).  There appears to be a bit of a fad happening at the cafeteria; everyone is starting to eat this soy-based yogurt stuff.  It's quite funny to see so many people with it on their trays all of a sudden.  It's 100円, and I've been getting hungry in the afternoon lately, so I bought some to try when I saw it in Yokosan today.  It's a little dessert-ey, so maybe not a good addition to lunch.  I will have to translate the nutritional information before I pass final judgement though.  I didn't play handiball after eating, as there was only Tsutsumida-san and I available today, and that's not really enough.

In the afternoon I puzzled away at how to fill in the deficiencies I think exist in the models I'm to implement.  The problem is that my current topic of study is online learning, which means the model is able to learn from new data as it comes in.  While this is cool, not to mention practical, it also means that the hardcore principled mathematical justifications behind all the formulas kinda get... flushed down. >_<  At present it seems a bit more like wizardry than anything, as we are putting decay-causing parameters on things we want to decay, but still are saying "this parameter is drawn from an Dirichlet distribution", which makes no mention of decay.  Iwata-san's paper seems to have a more principled approach, but it's far beyond what I'm working with at the moment.  I don't seriously mind the black-magicking that's going on, other than that it makes it a little hard to decide what to do when I believe I'm missing information, since I can't really derive it per se.

At the very end of the day I had an idea that I very much wanted to try, but it unfortunately would have taken at least 2 hours, so I couldn't stay to do it.  Oh well, at least I will have something additional to bring to my question session with Kabutoya-san. >_<  It would have been nice to have had the chance to test it out instead and just move on, but that's just how it goes sometimes. :|

I walked home, which was pleasant, but some sort of rather large bug seems to be... in bloom.  Or something.  I'm not sure how to describe it.  There are these carcasses of large, winged bugs lying around more often at the moment.  They aren't EVERYWHERE, but I had never seen one prior to yesterday, when I saw one or two, and on the way home I saw maybe 3 more, and then there was one more on the dorm stairs. :S  I'm hoping that bugs vs. time is not a linear function... Oh the humanity! ;D

Dorm dinner was pretty good.  There was a pork and garden salad, a piece of salmon and stir-fried vegetables with miso soup and rice.  After, I ate a grapefruit (my teeth seem to be better again, spoke too soon and hope they stay this way) and some of my multitudinous carrots.  After I just relaxed and unwound.  This week has felt like the longest yet by a long shot.  It's really hard to believe I went to Kyoto just 4 days ago. O_O  I had a nice text chat with Mom, and am now going to sleep until I wake up naturally, which will be a wonderful, wonderful experience.

Night!