sample_k provides sets of low volume

[jilljenn]

While it was supposed to do the contrary, right? Sets of higher volume are supposed to be more frequent.

I made a simulation where I ordered all subsets by decreasing determinant value, and the subsets sampled by sample_k seem low:

[10 13 15 16] det: 0.000274596023484 rank: 4837 out of 4845
[1 3 5 7] det: 0.412192537684 rank: 4349 out of 4845
[ 1  2  7 11] det: 0.0780018515433 rank: 4685 out of 4845
[ 1  5  7 17] det: 0.422286852875 rank: 4344 out of 4845
[0 5 7 8] det: 0.304245499402 rank: 4429 out of 4845
[ 3 10 16 19] det: 0.00256373499849 rank: 4832 out of 4845
[ 1  2  3 19] det: 0.00322178309274 rank: 4831 out of 4845
[ 1  7 10 12] det: 0.449683923531 rank: 4327 out of 4845
[10 13 15 16] det: 0.000274596023484 rank: 4837 out of 4845
[ 5  7 10 16] det: 3.76594172892e-05 rank: 4839 out of 4845