Idea for modeling categories

Why not set up a 12-dimensional lattice, in which each dimension maps
to one interval class out of 12, and then find the "temperaments" in
this lattice with the lowest "error?"

In this case, the val that has 0 error will be <1 2 3 4 5 6 7 8 9 10
11 12|, assuming that the coefficients map to 1\12, 2\12, etc. So we
want things that don't have 0 error, but slightly higher error.

It would be interesting to see if there's some kind of "uncanny
valley" effect that results.

-Mike