# [clug] Anyone want to give a talk for this week's PSIG meeting?

Ivan Lazar Miljenovic ivan.miljenovic at gmail.com
Tue Oct 12 00:31:15 MDT 2010

```On 12 October 2010 17:10, Kevin Pulo <kev at pulo.com.au> wrote:
> On Tue, Oct 12, 2010 at 11:35:12AM +1100, Ivan Lazar Miljenovic wrote:
>
>> How do you make a set of vertices and a set of pair-wise relationships
>> between those vertices "nice"? :p
>
> Simple.  Define a function f:(n)->(x,y) that maps each node to a point
> in the plane (or a point (x,y,z) in 3-space, if you think that'll
> help) and one g:(n1,n2,t)->(x,y) that maps each pair of related nodes
> and a parameter (between 0 and 1) to a point (such that g(n1,n2,0) =
> f(n1), g(n1,n2,1) = f(n2), and g(n1,n2,t) is continuously
> differentiable over t [0,1]), in such a way that you minimise
> w1*sum(h1(n,m)) + w2*sum(h2(n1,n2,m1,m2)) for all nodes
> n,m,n1,n2,m1,m2 such that n != m && n1 != m1 && n2 != m2, where
> h1(n,m) is defined as (f(n) == f(m) ? 1 : 0) and h2(n1,n2,m1,m2) is
> defined as the number of pairs (t1,t2) such that g(n1,n2,t1) ==
> g(m1,m2,t2), for given weights w1 and w2 - or you could choose to find
> suitable functions that maximise some other niceness metric, if you
> think that'd be better.  Disclaimer: part of my PhD involved graph
> drawing.

Oooohhh, care to elaborate? (graphviz has been annoying me lately)

>  Are we off-topic yet?  ;)

Let's wait until someone complains ;-)

--
Ivan Lazar Miljenovic
Ivan.Miljenovic at gmail.com
IvanMiljenovic.wordpress.com
```