On 9 Jan 2006, at 16:07, Oskar Sandberg wrote: > Ian Clarke wrote: >> One of the most useful parts of the visualization is a histogram >> of the link connection lengths. Some may recall that the goal is >> for the probably of the existence of a connection to be >> proportional to 1/ D where D is the length of the connection (ie. >> the difference between the node's locations). The effectiveness >> of Oskar's location swapping algorithm can be measured by how >> well the network conforms to this pattern. As you can see from >> the attached image, right now it is performing extremely well, >> even on such a small network. > > Not be the pessimist just because I am, but one thing that I have > seen is that the algorithm is very good at quickly creating a link > length distribution that looks correct. It is, unfortunately, a > much harder problem to make a network that routes well. (I can > create graphs with a great looking distribution, but where routing > fails horribly.)
How would you create a network with good-looking distribution, but which doesn't route well? Ian.
