Talk:Brownian tree

Note that "collide" is not defined. In one implementation "collide" means that the particle lands on a fixed particle (and then its previous position is used). In another implementation (mine), "collide" means that the particle reaches a position adjacent to a fixed particle. I do not know if this ambiguity matters, but I thought I would mention it. --Rdm 11:54, 14 July 2010 (UTC)

It's interesting, but I don't think it matters; there are enough differences between languages that a retrace step in the algorithm doesn't seem important. --Michael Mol 16:38, 14 July 2010 (UTC)

It does make a difference in the resulting tree: the adjacency strategy makes particles have a larger "cross section" for collisions, and it will probably cause a difference in the appearance of the tree. But an even stronger influence is where the particles are added to the field. —Kevin Reid 22:25, 19 July 2010 (UTC)

Given a specific sequence of numbers to have fed out of the PRNG call (or using the same seed each time), it would be interesting to compare outputs from the two approaches. --Michael Mol 15:53, 20 July 2010 (UTC)

The tree shape is very sensitive to how collision is defined. The way each new particle attaching to the tree affects who itself can be used as an anchor later, sort of like a fractal. From tests I did, small changes to some details (how far away the stuck particles are from each other, whether Brownian motion is slightly biased, at what distance are two particles considered colliding, etc) make the trees look completely different. Some code examples produced images that don't look like trees at all, I guess this is why. --Ledrug 02:55, 11 June 2011 (UTC)

C and D entries

Originally the D entry derived from the C entry. And later some other language entries have derived from the C/D entries. But the I think the C and C-derived entries don't respect the rules well (it moves already set particles, it doesn't always attach a particle if one of its Moore neighborhood cells is set), so I have rewritten the D entry. I think this D entry is closer to the literature value (1.71) of the resulting fractal dimension of DLA process. The fractal dimension is now lower than the C version (so the resulting tree is less dense), because the attach probability is higher, this causes a faster dendrite growth.--Bearophile 23:19, 27 June 2012 (UTC)