Saturday, June 30, 2012

Saturday: I probably should have done more work today.

Instead I went to get sushi for lunch.  I went to Genki, as I didn't really want to drive out to Kahala.  Genki does have nice labels that circle with the food, so you know exactly what you're getting.
 I did end up sitting next to possibly the rudest girl ever, which was annoying.  I'm pretty sure no one wants your victoria's secret bag sitting on the counter.  Also, waving and snapping your fingers at the sushi guy because you want, "Um...that sauce? You know? For tempura? Can I have some?" makes you look like an ass.
Their unagi isn't as good as Kuru Kuru.
 On the way out, I had to wait for some lady to pay her bill, then ask if she could exchange a hundred dollar bill for twenties, and then conspicuously put two dollars in the tip jar.  She did all this while the checkout guy looked awkwardly at me because neither of us could get my bill paid until she moved.
This karaage took fucking forever to come out.
 Did you know they make blood orange San Pellegrino?  It's better than the other flavors.  I picked it up while getting my spite foie gras at foodland.  Mmm...spite....
What are you doing dogs? You're not trained biohazard experts.  You shouldn't even be here.
 I also discovered today that making a Levenberg-Marquardt change to a fitting algorithm does increase the convergence speed.  "Duh," I hear you say.  Yes, but when I first implemented the fitting, I intentionally went with a standard non-linear least squares fit because the GSL LM solver would get trapped in local minima. I'm still not sure why my version doesn't have that problem, beyond the fact that I've changed a lot of the other calculation stuff.  In fact, adding the LM step was brought on by oscillatory solutions where the fits would bounce around the chi^2 minimum in a cyclical pattern.  I thought that adding something that changed the step interval each time would solve that.  It seems to do so, so that's good.
Random Charizard.

No comments:

Post a Comment