One may optimize the DP above with divide-and-conquer accompanied by Taylor Shift, to a total of $O(N^2log^2 N)$ time with a fairly large constant factor.

虽然我根本不可能学会但还是好奇这是什么神仙科技.jpg /kel

主要是那个 Taylor Shift，我必应到了一篇算是论文的东西？但看的也很迷惑： 这里。