[Networkit] Issue with Directed Graph and Question

Michael Marcucci onlycutch at gmail.com
Tue Oct 6 14:44:47 CEST 2015


At the moment I am at a conference and don't have the code with me. But it was something like 
g = graph(g, false, true)
for line in f:
    L1 = g.add_vertex(1)
    L2 = g.add_vertex(1)
    g.add_edge(L1, L2)

sc = properties.StronglyConnectedComponents(g)
sc.run()
Fails here output
Bus Error (Core Dumped)

I can try and send the core when i get back as well, haven't yet had the time to get it. 

As for everything else thank you that i very helpful. 

Michael

> On Oct 6, 2015, at 4:51 AM, Christian Staudt <christian.staudt at kit.edu> wrote:
> 
> Hi Michael,
> 
>> On 05 Oct 2015, at 17:50, Michael Marcucci <onlycutch at gmail.com> wrote:
>> 
>> First off great program works very nicely, and does not eat up my RAM.
>> However I did have an issue running properties.StronglyConnectedComponents(g) on very large directed graphs.
>> It fails on Ubuntu x64 with a Bus Error.
>> My graph is 800,000 vertices & 135M edges.
>> As a test of the software I pretended they were un-directed edges and the ConnectedComponents function worked fine.
>> The machine has 160GB of RAM and only ~25GB are being used at the time of the failure so it is not a lack of resource issue.
> 
> How exactly does the failure look like? Can you show the code/output?
> 
>> 
>> As for my question, the Betweenness and Closeness algorithms are they suitable for directed graphs? Most of the algorithms I have seen only use un-directed graphs. Is the Networkit algorithm doing something different, or accounting for directed-ness?
> 
> Here’s the policy: Any algorithm class that doesn’t work with a particular type of graph throws an exception from the constructor. It is possible for the class to check for directedness and do something different in that case.
> (In general I recommend one class <-> one measure rather than one class <-> one algorithm, i.e. a class can contain multiple algorithms and run them depending on the input)
> 
> As for Betwenness and Closeness, they rely on the BFS and Dijkstra classes which compute shortest paths and handle directedness transparently.
> 
> Best,
> Chris
> 
> 
> _______________________________________________
> NetworKit mailing list
> NetworKit at ira.uni-karlsruhe.de
> https://lists.ira.uni-karlsruhe.de/mailman/listinfo/networkit



More information about the NetworKit mailing list