Talk:Random Latin squares: Difference between revisions

No edit summary
Line 1:
==[[Latin Squares in reduced form]]==
[[Latin Squares in reduced form]] makes trivial generating uniformly generatingdistributed random Latin Squares up to order 6. 3 random numbers are required. The first in the range 1 to number of reduced Latin Squares of order n is used to select a member of the set of reduced Latin Squares of order n. The second in the range 1 to n! is used to select a permutation of the columns of the selected reduced Latin Square. The third in the range 1 to (n-1)! is used to select a permutation of the rows 2 to n.--[[User:Nigel Galloway|Nigel Galloway]] ([[User talk:Nigel Galloway|talk]]) 09:51, 12 July 2019 (UTC)
===Random Latin Squares of order greater than 6==
This task becomes interesting at order 7. Consider:
<pre>
1 2 3 4 5 6 7
2
3
4
5
6
7
</pre>
There are 5! ways to add 7 to the above. Having selected one of them it is necessary to determine the number of ways of completing the Latin Square similar to the way used in [[Latin Squares in reduced form]]. Selecting one of the completions and permuting the columns and rows as above will produce an almost uniformly distributed random Latin Square. The distribution is now not quite uniform because the number of ways of completing the Latin Square will vary with the arrangement of 7s chosen.--[[User:Nigel Galloway|Nigel Galloway]] ([[User talk:Nigel Galloway|talk]]) 10:06, 12 July 2019 (UTC)
 
==Revert to draft==
I think the task description needs to require and demonstrate that the algorithm must be capable of producing all valid latin squares of size n. As it stands starting from a valid latin square, which is easy to generate say:
2,172

edits