Re: [igraph] R-implementations of games

[Arthur Kaiser]

Hi,
@Tamas. Thanks. I want to do it right this way.

@Gabor

How can I get and use an edge list matrix? Can you give me a code 
example, please?

I'm actualy thinking about an aging.prefatt.game with a random part:

P[i] ~ (c k[i]^alpha) (d l[i]^beta) (e r[i]^gamma)


The combine which I´m referring is shown in this example:

my.example.game<-function(n, directed=TRUE){

       graph<-graph.empty(directed)
       graph<-add.vertices(graph, 1) #add first vertex


    for(i in 1:n){

        # and the rest
        graph<-add.vertices(graph, 1) #add a vertex

        #regular latice
        graph<-add.edges(graph, c(i,i-1))
        graph<-add.edges(graph, c(i,i-2))

        #some flavour of er
        graph<-add.edges(graph, c(i,sample(1:i-1, 1)))

    #some flavour of ba
        startvertices<-graph[4][[1]] #incoming ends vector
        maxj<-length(startvertices)

        j<-sample(1:maxj, 1)
        vj<-startvertices[j] #to

        if(maxj==0){graph<-add.edges(graph, c(i,0))# add first edge
        }else{graph<-add.edges(graph, c(i, vj))}

    graph<-simplify(graph)
    }
return(graph)
}

I don´t think that this would be an usefull combination. I like to try 
out some -more or less strange- ideas, especially quick and easy ones. 
You are right. R implementations are slower, but performance is not the 
problem. It's ok for me to wait some seconds longer.
If I create the union of two graphs (e.g. ba+lattice) the result should 
have different properties than the graphs which I generated within my 
example (without random part). What do you think about this?

Best Regards,
Arthur

Am 02.02.2010 13:29, schrieb Gábor Csárdi:
> Hi,
>
> On Mon, Feb 1, 2010 at 8:58 PM, Arthur Kaiser<address@hidden>  wrote:
>    
> > Hallo,
> > do R-implementations of games exist?
> >      
> no, at least not within the package. There might be some code
> available here and there, e.g. on this mailing list.
>
>    
> > I especially need a R-implementation of aging.prefatt.game.
> > Here is an example of a simple barabasi.game implementation:
> >
> > my.ba.game<-function(n, directed){
> >         graph<-graph.empty(directed)
> >         graph<-add.vertices(graph, 1) #add first vertex
> >
> >
> >     for(i in 1:n){
> >         # snapshot
> >         # if(i==x) {write graph}
> >
> >         # and the rest
> >         graph<-add.vertices(graph, 1) #add a vertex
> >
> >         startvertices<-graph[4][[1]] #in degree
> >      
> These are indeed the "incoming ends" of the edges, but I would not do
> this. This is a violation of the API. Why don't you simply use an edge
> list matrix in this loop (probably with pre-allocated size to save
> time)? You are not really using anything igraph specific, anyway.
>
>    
> >         maxj<-length(startvertices)
> >
> >         j<-sample(1:maxj, 1)
> >         vj<-startvertices[j] #to
> >
> >         if(maxj==0){graph<-add.edges(graph, c(i,0))
> >         }else{graph<-add.edges(graph, c(i, vj))}
> >
> >     graph<-simplify(graph)
> >     }
> > return(graph)
> > }
> >
> > Are you wondering why I ask for this? I need snapshots of the graph during
> > the generation process. I also want to combine games e.g. ba with a d-dim
> > latice or a randoom game without writing it in c and compile. Can someone
> > help me please?
> >      
> Tamás already told you about the snapshots. I'm not sure what you mean
> by 'combine', but you could just create the lattice separately and
> create the union of the two graphs. See e.g. graph.union.
>
> In general, we don't want to rewrite everything in R, just to have
> another implementation, so I'm afraid that you'll need to do this for
> yourself. Note that the R implementations of these functions will
> typically contain 'for' loops and they will be much slower than the C
> versions.
>
> Best Regards,
> Gabor
>
>    
> > Thanks in advance.
> >
> > Arthur Kaiser
> >
> > University of Applied Sciences Brandenburg
> > Faculty of Computer Science and Media
> >
> >
> > _______________________________________________
> > igraph-help mailing list
> > address@hidden
> > http://lists.nongnu.org/mailman/listinfo/igraph-help
> >
> >      
>
>
>