allentran / graph2vec Goto Github PK

https://github.com/benedekrozemberczki/karateclub

Karate Club is an unsupervised machine learning extension library for NetworkX.

Karate Club consists of state-of-the-art methods to do unsupervised learning on graph structured data. To put it simply it is a Swiss Army knife for small-scale graph mining research. First, it provides network embedding techniques at the node and graph level. Second, it includes a variety of overlapping and non-overlapping community detection methods. Implemented methods cover a wide range of network science (NetSci, Complenet), data mining (ICDM, CIKM, KDD), artificial intelligence (AAAI, IJCAI) and machine learning (NeurIPS, ICML, ICLR) conferences, workshops, and pieces from prominent journals.

When using parse_graph(), I get a file/directory not found that points to save_mappings() in parser.py. Lines 47 and 49 point to nonexistent file/dir, but this problem persists even if I set the output_dir in creating the Graph2Vec object. I think it's just a matter of addressing the os.path.join(), but not sure. I can have a look later today and submit a pull request if that would help.

See the traceback below. This doesn't work on Ubuntu 14.04 (AWS GPU instance) running Python 2.7.6.

`In [4]: graph2vec.parse_graph('test.txt', extend_paths=2)

IOError Traceback (most recent call last)
in ()
----> 1 graph2vec.parse_graph('test.txt', extend_paths=2)

/usr/local/lib/python2.7/dist-packages/graph2vec/trainer.pyc in parse_graph(self, graph_path, data_dir, load_edges, extend_paths)
26 graph = parser.Graph(graph_path)
27 self.from_nodes, self.to_nodes = graph.get_mappings()
---> 28 graph.save_mappings(self.output_dir)
29
30 if load_edges:

/usr/local/lib/python2.7/dist-packages/graph2vec/parser.pyc in save_mappings(self, output_dir)
45 def save_mappings(self, output_dir):
46
---> 47 with open(os.path.join(output_dir, 'from.map'), 'w') as from_map_file:
48 json.dump(self.from_nodes_mapping, from_map_file)
49 with open(os.path.join(output_dir, 'to.map'), 'w') as to_map_file:

IOError: [Errno 2] No such file or directory: 'data/from.map'`

I installed graph2vec successfully but I cannot use it

import graph2vec.trainer
graph2vec = Graph2Vec(vector_dimensions=128)
Traceback (most recent call last):
File "", line 1, in
NameError: name 'Graph2Vec' is not defined

Hey, I'm trying to compare separate graph vectors to each other -- do you have to train each one independently? Would that result in the same vector space representation? How could somebody learn on a bunch of different graph structures? I didn't see a way to do this in the Readme, and didn't see how to do it in the source.

Thanks.

Hi,

Thanks a lot for the implementation of graph to vector method. I am vaguely familiar with several methods, e.g. Laplacian Eigenmaps and similar methods. The implemented method in this module looks like is matrix factorization style algorithm (?) with square loss (?).

It would be great just to add a reference to a paper/blog which actually describes the method with equations.

Thanks

I failed to use graph2vec with python 3.x
A patch is provided:
sed -i 's/cPickle/pickle/g;s/from node_vectors import/from graph2vec.node_vectors import/g' graph2vec-0.0.2/graph2vec/*.py