Consider the unit disk $C \subseteq \mathbb R^2$, and imagine the following process:
Repeat $N$ times:
- Take a strand of spaghetti (the length is fixed!) and randomly place it in the disk so that it forms a smooth curve with the 2 endpoints pined on at points on the circumference $\partial C$.
It's required that as $N \rightarrow \infty$, the distribution of spaghetti material in $C$ becomes uniform in one of the following senses:
A) Length: For every subregion $S$ of $C$, the ratio total length within $S$ to the total length within $C$ uniformly approaches the ratio of area of $S$ to area of $C$.
B) Particle count: For every subregion $S$ of $C$, the ratio of the total amount of spaghetti material within $S$ to the total amount of spaghetti material within $C$ uniformly approaches the ratio of area of $S$ to area of $C$, where we define the density of spaghetti material at a point to be the number of spaghetti strands passing through that point.
Question
- How would the sampling be done so as to achieve this?
- Can one succinctly characterize all the possible sampling procedures that can achieve this ?
- How about making the convergence to uniformity as rapid as possible.