On 23.02.2015 23:30, mohsenuss91 wrote: > Hi, I'm working on graph mining, so I'm trying to find the best library to do > that. > I read in http://graph-tool.skewed.de/performance that "graph-tool" is > faster, so I tried the same program who count the duplicated graphs (I call > them frequent in the program) in networkx and graph-tool. > graph are in this .txt file ==> http://pastebin.com/u5BUxx2p > networkx program ==> http://pastebin.com/FPRy7Ywv > graph-tool program ==> http://pastebin.com/dK93x6tf > > > there are the results: > * Networkx: > RUNNING TIME: 0.00204300880432s > Reading from file: 0.000211000442505s > Algo: 0.00186109542847s > > > ************* > graph-tool: > RUNNING TIME: 0.0780489444733s > Reading from file: 0.00203204154968s > Algo: 0.0760469436646s* > > ___________________________ > I 'm not sure if my graph-tool program need amelioration or that is the best > performance ?
These differences are interesting. I expect graph-tool to become faster if the graphs become larger (your graphs are very small). Networkx compares the degree sequence (and the vertex invariants) before running the actual isomorphism code. This is a shortcut that improves things if the graphs are clearly not isomorphic. Do you know if this is the case for your graphs? What happens to the time difference if all the graphs tested are isomorphic? I can easily implement these shortcuts in graph-tool as well, if they happen to be the source of the discrepancy. Best, Tiago -- Tiago de Paula Peixoto <[email protected]>
signature.asc
Description: OpenPGP digital signature
_______________________________________________ graph-tool mailing list [email protected] http://lists.skewed.de/mailman/listinfo/graph-tool
