1
$\begingroup$

I have a connected, sparse, graph (a molecule to be specific) and I'm interested in associating 3D coordinates with the vertices. Here's the kicker: I already have coordinates for none/some/all vertices which are not to be changed. The case of no coordinates I can solve with e.g. MDS, and the case where I have all coordinates is trivial.

In addition, I have a length associated with every edge which should be respected (but some distortion is fine). If it helps, I also have information about some (not all) angles and dihedral angles. Finally, all vertices should be separated as much as possible from one another. If need be, you may assume a maximum degree of ~6.

I don't expect a fully worked solution, but any pointers to literature would already be greatly appreciated!

$\endgroup$
4
  • 1
    $\begingroup$ Maybe write down the energy function and perform simulated annealing? $\endgroup$ Commented Jul 11, 2019 at 15:55
  • $\begingroup$ Thanks for the input, I'll give that a try. Although it feels somewhat unsatisfactory ;) $\endgroup$
    – pckroon
    Commented Jul 12, 2019 at 13:25
  • $\begingroup$ There're lots of optimization algorithms to try, not only limited to simulated annealing. $\endgroup$ Commented Jul 12, 2019 at 16:22
  • $\begingroup$ Don't worry, I realize. The first thing I'll try will be some local optimization, preferably one I can take from scipy. But that's more StackOverflow than MathOverflow. $\endgroup$
    – pckroon
    Commented Jul 16, 2019 at 9:41

1 Answer 1

1
$\begingroup$

There are a number of papers by Mike Treacy (I. Rivin is a co-author on some) which address this problem, but in a purely practical manner, using essentially the scheme proposed by Bullet51. Here is one:

Enumeration of periodic tetrahedral frameworks
M Treacy, KH Randall, S Rao, JA Perry, DJ Chadi
Zeitschrift fur Kristallographie 212 (11), 768-791
$\endgroup$

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .