Is there any formula to produce a coherent noise, where it is guaranteed that no 2 unique values will be present inside any subset of size s ?
Asked
Active
Viewed 42 times
0
Dávid Tóth
- 133
-
Can you clarify what you mean, perhaps by example? Are there are "bins" and the noise can appear only once in a bin? – user619894 Aug 04 '21 at 12:41
-
@user619894 thanks for asking! I can provide program code which generates an eligible sample. The logic is the following: for each point in the target space a random value is set; Then The logic checks for each point if any in-range points have a similar value, and if it does the value of the cell is modified to another random value. It is a greedy way of generating it, but it works. I can Detail these steps in the answer if that would help. – Dávid Tóth Aug 04 '21 at 18:23
-
If the rng has a repeat length much longer than the number of points, this should be very rare event, even for very distant points. In fact, it is probably easier to generate streams of globally unique pseudo randoms than to specifically demand them to be distinct in regions. – user619894 Aug 05 '21 at 05:57
-
Indeed it is a rare occurrence, but I believe there should be a mathematical formulae to satisfy such requirements – Dávid Tóth Aug 05 '21 at 06:06
-
A standard rng has no notion of distance. You can try casting out all duplicates before applying , or maybe some trick with hash tables. As a physical model this sounds like a large q Potts model with antiferromagnetic interactions: nearby neighbors don't like to have the same q value. maybe there is something relevant in the physics literature. – user619894 Aug 05 '21 at 06:15