I tried something different now. While it's not as beautiful as ztanz, it's not too far away of that in my opinion:
First, I tried to do a logistic growth Mset. Not too interesting. The graph already made that to be expected.
Then I found out, that certain logistic growths simplify to 1+tanh(z) - still not interesting, but the connection to tanh(z) made me even more curious about the results. - It now seems like tanh(z) is in a way the base of any logistic growth.
Then I thought: Hmmm: this thing actually looks like an integral. It's how many individuals exist per timestep.
So, my next idea was, to take the differential of tanh(z) which is sech²(z). - It's basically how many individuals become added in that certain time-step.
So that's what I'm rendering right now. - The anti-buddhabrot version for now.
It looks very nice and promising. Still quite noisy though. 1/sin(iz)² - I just renormed it a little bit so that the integral over the total set gives 1 but that shouldn't change too much. In theory, that curve now could be used as a probabillity function, at least over the reals
So, the actual formula I'm using is
with a being any positive real - they will always produce 1 as an integral. The specific value of a, I chose, was
, as that made it more symmetric... (2/sqrt(2) = sqrt(2))
Probably, you could directly use sech²(z) for nice results, too. The image shouldn't vary too much...
Oh, btw, I also tried the exiting orbit version. It looked nice as well but cleared up very slowly...
There surely are good reasons for using the gaussean normal distribution but the sech(x)² shares the property of the simple integral and on the imaginaryaxis it's pretty much like a cyclic version of the normal distribution... Looking at the general normal distrubution with sigma and µ, I think it would be very trivial to add that to the sech distribution aswell...
EDIT: Ok, found. It's actually called Logistic distribution (d'uh)
http://en.wikipedia.org/wiki/Logistic_distributionEDIT:
Argh! As usual: Errors make beauty. I missed an h, turning sinh to sin. Now I have to check, wether it still looks as nice, without the error...
EDIT:
Lucky: It changed shape quite a bit, but it still looks interesting
(Hyper short rendertime 'till now - just a few secs... But it already looks reasonable) - certain features still are more or less unaltered, actually...
If you want to look at the error, just use the above mentioned formulae but change the line
to