SNAP: Stanford Network Analysis Project | Hacker News Reader

SNAP: Stanford Network Analysis Project

snap.stanford.edu

61 pointsindescions_201712 comments

What I like most about this project is actually the repository of data sets. Curated graph data sets are hard to come, and the ones at SNAP have already been used in different publications, so there are fewer errors and, usually, a good baseline for comparisons available. Plus, the data sets are relatively diverse.

It is really good at performance (thanks to optimised C++ implementation) for running it on large networks compared to networkx or other pure python implementations but its usability is very bad.

Agreed. Useful but documentation is minimal, uses some non-standard naming conventions.

Looking for a similar library that works with Python 3? Checkout NetworkX. It's awesome! http://networkx.github.io

Military funded (DARPA's Social Media in Strategic Communication program) to develop the research necessary for running highly automated global-scale social network propaganda campaigns.

Can someone explain why this is coming up now? From what I see, this looks like an old project (2014). I'd love to understand if there is something fresh to be aware of.

Well, I didn't know about the project, and now I do. Looks really promising, too.

EDIT: except, looks like they don't support python 3 yet. That's a dealbreaker, unfortunately. All my code is in python 3.6.

I don't care about a bunch of Python 2 research-quality code. The amazing part of the project, to me, is all the data that they make available in one place.

Another great source of network data: https://icon.colorado.edu/#!/

If you are interested in data, maybe KONECT (http://konect.uni-koblenz.de) is also for you. Or try the benchmarks from Dortmund University (https://ls11-www.cs.tu-dortmund.de/staff/morris/graphkerneld...).

I have used SNAP because it implements a role extraction algorithm for graphs. But in my experience networkx, python-igraph and graph-tool are nicer libraries to use for manipulating graphs and performing network analysis. Networkx is perhaps the most Pythonic, but it's also the slowest.

https://networkx.github.io/ http://igraph.org/python/ https://graph-tool.skewed.de/

It's even older than that. I remember it from back in grad school (ca. 2009) when Facebook and social networks in general were hot subjects in academia.

Reply on news.ycombinator.com