2002-12-05 11:54 pm (UTC)
good against "all the statistical tests in Knuth"
Sparking heated discussion in the Q-and-A after a Diaconis donuts on statistical tests of PRNGs. He was incensed when Diaconis remarked something about bothering to run the statistical tests on Blum-Blum-Shub (I love saying that), when the theory says it's (asymptotically) just got to pass.
Yeah, that "asymptotically" sounds like an important word there.
Look at it early.
Complexity theory is not (as I found out) one of those "cram 48 hours of take-home work into the last two 24 hour periods before its due" classes.
For "Image-Based Modelling and Rendering", on the other hand, this approach was a stunning success.
Well, I mean, I got a "B" in the class, but I think I've got a good reccomendation out of Seitz based on my final project.
Amazing how two classes with the same number prefix (8xx) can differ by orders of magnitude in terms of difficulty (I like to tell myself that there were big gaping holes in my logic and computability background which, while only tangentially related to complexity theory, would've greatly helped out, but really, there were also big gaping holes in my vision background [read: no vision background at all] when I took IBMR, and they didn't seem to hinder anything)