From @ZERMATT.LCS.MIT.EDU:sandy@MEDIA-LAB.MEDIA.MIT.EDU Sat Mar 25 12:02:42 1989
Subject: oops..
To: hkuo@media-lab.media.mit.edu

I think I suggested the wrong energy function;
what we really want is something like
 (abs (* (sin theta) (cos theta)))

where theta = (acos (dot-product v1 v2)), as this
has zeros at 0, 90, 180, and 270 degrees.

   Sandy

From sandy@media-lab.media.mit.edu Fri Mar 24 09:47:54 1989
Subject: Re:  ARVO
To: hkuo@WHITECHAPEL.MEDIA.MIT.EDU, adelson@UXBRIDGE.MEDIA.MIT.EDU

I think the problem may be that straight lines
are OK too, so that we need an external energy function
that has two minima: one where the lines are at 90 deg,
and one where they are at 45 deg.  

An exaple is (cos (* 2 (acos (dot-product v1 v2))))
where dot-product normalizes the length of vectors
v1 and v2.

  Sandy

