JPF wrote:I finally managed to find my old programs.
If the question is still relevant, I confirm the numbers provided by Blue.
Some elements for their calculation according to the method explained
here:
Number of permutations: 2 x 6^6 = 93312
Number of conjugacy classes : 405
Class number, size, cycle decomposition:
(...)
Possible cycle lengths = list of i such that m_i is non-zero: 1,2,3,4,6,12
To Blue: Which method did you use?
Hi
JPF,
Thank you very much, and special thanks for the link to the explanation.
The question was still relevant for me.
It's was work than I imagined, doing the conjugacy classes and all, so thanks again.
In principle, I had generated one shape from each equivalence class, using a canonical form that's equivalent to the "minlex" version of the shape, when the cell data is read out in the order shown here:
- Code: Select all
+----------+----------+-------+
| 28 29 30 | 46 47 48 | 1 2 3 |
| 31 32 33 | 49 50 51 | 4 5 6 |
| 34 35 36 | 52 53 54 | 7 8 9 |
+----------+----------+-------+
| 37 38 39 | 19 20 21 | |
| 40 41 42 | 22 23 24 | |
| 43 44 45 | 25 26 27 | |
+----------+----------+-------+
| 10 13 16 | | |
| 11 14 17 | | |
| 12 15 18 | | |
+----------+----------+-------+
It's highly "box oriented".
Taking every shortcut I could, I ended up generating ((36*35/2)*36 + 36*26)*512*512*512 = 3,169,685,864,448 test cases, and checking them for being canonical forms.
It was a ratio of ~7.33 test cases per actual canonical/minlex form.
[ 512 = # of box fills ]
[ 36 = # that are ED w/rsp to row & column perms ]
[ 26 = # that are ED w/rsp to diagonal reflection and row & column perms ]
The Wikipedia article on Burnside's Lemma, mentions using "enumeration/generation" with "minlex forms" as a way of verifying that "Burnside's lemma was correctly applied".
I was hoping for the opposite: to use a calculation from you or from Serg, as a check that I didn't take too many shortcuts, and (likely) didn't have bug(s) in by canonicalization code.
That and I thought it might be a fun exercise for you guys !