Jason (jcreed) wrote,

Dang, I have actually been kind of good at being social lately. Today I

  • hung out for quite a while at the food TG, which featured Chick-Fil-A chicken nuggets. They were slightly cold, but still delicious.
  • ran into cute girl from Horse and Burro Adoption WRCT thing, did not shrink from the opportunity to say hi again.
  • Went to nolacoaster's party, met some new people, enjoyed talking to old people. Really I think the only reason I left is I was getting tired of standing, and didn't feel like consigning myself to the couch-room.

In the background I have been chugging away at several projects, thesis work included among them. However, thesis work does not generate pretty pictures! Other projects do. I made some caps for "Tragic Sans":

Also I was suddely inspired a few days ago to recreate some code I think I originally wrote around sophomore year while on vacation in montana visiting my (late) grandparents. The idea was to do something like heat diffusion, but over a nonstandard notion of "temperature". It would ordinarily be just repeated blurring, except you make the value at each pixel not a real number, but a point on the circle S1. I faked this by having every point keep two real numbers, and at each iteration doing a blur and renormalizing back to unit length. I wouldn't be at all surprised if this has physical significance somewhere, (alignment of magnetization or something, maybe?) but I'm ignorant enough about physics that I don't know what it would be.

The cool thing about this is that because S1's fundamental group isn't trivial, you get little poles which the diffusion function can't "blur away", because the values make a complete round-trip around S1. However, since some of them are clockwise, and some are counterclockwise, they can collide into one another, cancelling out. I suppose it ought to be possible to get other elements of the funamental group, but I haven't seen them appear out of noise. Here are some static images from the program:

blue and green channels convey value in S1, red channel displays a time-derivative. The little red wavelets appear where "particles" and "anti-particles" recently collided.

Here the green channel is just a thresholded version of value in S1.

Here the green channel is displaying the magnitude of some kind of curvature tensor, which tends to get very large and wrap around when a pole is nearby.

This is just a composite of 2 and 3 that I think looks neat.
Tags: fonts, hacking, social

  • Post a new comment


    Anonymous comments are disabled in this journal

    default userpic

    Your reply will be screened

    Your IP address will be recorded