Re: [igraph] Graph Visualization

[Gabor Csardi]

Lorenzo,

first of all. Igraph is not very good at visualization, so you 
might want to look for other options if you'd like something fancy.

As for your problem, if your particles live in the 2d or 3d space, 
then why don't you just use their real coordinates? It might 
be a good idea to have the same color for vertices that are in the 
same component, although it is quite impossible to have 70 or more
distinguishable colors. 

Coloring vertices according to degree can be also good, but if you use 
the colors for the different components, then you can still use
different circle/sphere sizes for the vertices according to their degree.

Oooops, i just realized that you want to do this for 5000 vertices. 
Well, that won't be easy. The best would be to use a tool that can 
aggregate densely connected vertices into a single node and allows 
you to zoom in and out. I've seen such a tool presented by one of the 
GraphViz guys, but i cannot find it on their web page, so it is not 
out yet I think. (I've seen it about a year ago.) Prefuse and Jung 
are also great tools, but you'd require a lot of programming to use 
them. Another nice one is touchgraph, but that is not open source.
In biology many people use Cytoscape, i'm not really familiar with it,
i don't know what it can do.

Even generating a decent layout for 5000 vertices can be difficult,
if the graph is sparse, then the Fruchterman-Reingold algorithm
should be able to do it, but might also want to check DrL:
http://www.cs.sandia.gov/~smartin/software.html

Best,
Gabor

On Fri, Jun 06, 2008 at 06:50:58PM +0200, Lorenzo Isella wrote:
> Dear All,
> I would like a piece of advice about how to visualize graphs. I know
> very little about this; I simply browsed through the igraph site. I
> understood that different conditions may make a certain visualization
> technique more or less appropriate.
> This is what I am using igraph for and what I aim at:
> (1) I study the aggregation of particles which form large clusters by
> sticking together. I have a time-dependent adjacency matrix, which I
> save as time progresses.
> (2)I look for the connected components of the non-directed graph.
> Initially, I have 5000 distinguished connected components (each
> consisting of a single vertex; each particle is a vertex)
> (3) As time progresses, I have fewer and fewer connected components
> since most of the particles end up in a few large clusters. At the
> end, I have about 70 connected components.
> (4) I would like to take several "snapshots" of the graph to visualize
> what is going on, but I am quite unsure about what I should use, if it
> is possible to have different colors for each node depending on its
> degree and so on.
> 
> Any suggestion by the experienced people on this list is appreciated.
> Kind Regards
> 
> Lorenzo
> 
> 
> _______________________________________________
> igraph-help mailing list
> address@hidden
> http://lists.nongnu.org/mailman/listinfo/igraph-help

-- 
Csardi Gabor <address@hidden>    UNIL DGM