Would a split-the-difference approach help any? It yields a symmetric matrix at the midpoint of the element pairs that have any difference. If it's any consolation, it looks like the elementwise differences are down around 10 -20 or so.
0 kommentar(er)
