quick notes

Bought groceries today, including a now-empty box of kolacky. This is probably not good for me, especially since today was a normal day.

I don't get any exercise on normal days.

Tried some variations on the maptask problem, including a hack that added label estimates based on the previous element of a data sequence and a transition matrix. (It failed.) And binarizing count features. (Failed.) And adding a feature that was the number of unique words in an utterance. (Failed.) Well, 'failed' is harsh, since I'm using it in the sense of 'did not improve results'.

I'm not taking into account the fact that the label classes are of different sizes, e.g. we expect to find differing numbers of checks and clarifications and binary questions, etc. Future plan: implement Abe-Zadrozny-Langford method for reducing this weighted multiclass classification problem to binary classification.

Looking at SVMStruct now, specifically the files Thorsten set up to add new SVM struct objects. I so don't want to do this. I need some endorphins. It's too cold to run, and even if it wasn't, I'd find some other excuse.

Discovered SPIDER. Cool! And they need more visualization components. Woohoo, more stuff to do...

Did some complicated googling to find a draft paper that probably wasn't meant to be found. Will resist linking to it.

Discussed showers with Mark. Finally, someone testing my documentation! :) He wants a web interface, so that someone can fill out a form defining the initial cosmic ray / particle and then get back a VRML model of the shower structure. And he suggested replacing points by lines. I wish I remembered properly what went wrong the last time I tried doing that. In principle, it's a good idea, but ... what happened? ... were the lines too small? Trouble is, I can make thick lines in partiview now, so I can actually add this change...

Ah, the temptation to focus on what I can do rather than what I should do... trouble is, whenever I try to do the reverse, I do neither what I can do nor what I should do. [Think : empty box of kolacky.]

Sent off an email to the Physics Education journal asking if they were interested in an article on visualizing cosmic ray showers. They had better be. And to a second GRC Ed/Vis organizer to ask when the poster deadline was. The first organizer I emailed a week ago didn't bother replying.

Had a pleasant surprise from a visualization professor who found my site and invited me to give a talk if I was ever in his neighborhood. And it's a great neighborhood, too! Pity I won't be going there in the near future...