Thank you, Reuben! This platform does look very promising but the The installation is proving painful. I haven't found a straightforward way to download and install Octave.
I will have to fiddle with it until I manage to get it done.
What version of R do you have? What it is checking for is the version and if 3.2.x is not found it will fail. I think the new version of R is 3.3 and this will fail.
I also came across this one and played with it. I got all the examples on the website to run. The only problem that this one doesn't support CSV or textual data. It only works with matrices. Thanks!
You can remove 3.3.0 and install 3.2.5 or any 3.2.x version. Only do this if you dont need 3.3.0 and you dont have existing code that depends on the newer version 3.3.0.
Do you mean "geographical" networks such as road or railroad networks? If so, you could have a look to OpenMap (http://openmap-java.org). Among the different useable formats, you can find CSV layers.
Do you mean neural networks? It isn't so clear from the question.
Anyhow, in Python there exists a brilliant package Pandas with native support for .csv import. I believe it should then be possible for you to transform data into matrices and use the Pymnet package that Tanguy Fardet proposed.
Note: iGraph has a python version too... (third link)
From all above my favourite has been NetworkX in Python, so far. I assume your *.csv file contains edge pairs. Let us know if you need further help. Hope this helps...