search

williamleif / GraphSAGE

Representation learning on large graphs using stochastic graph convolutions.
Other
3.41k stars 841 forks source link

How to generate the files in example_data #54

Open ajosemf opened 5 years ago

ajosemf commented 5 years ago

Hi,

I downloaded a dataset and, from it, I need to create the files in example_data folder. Should I create my own code to generate it or is there code ready for this?

Tanks in advance

RexYing commented 5 years ago

Hi,

Because each dataset has a different format. We try to process all dataset into the same format to be used in the model. So you need to write custom code to create the input files in example_data folder, depending on the actual format of the graph dataset you want to use. The input are just features, adjacency etc. in the form of numpy matrices, so it should be straightforward to generate.

You can refer to the input format section of readme.

For the walks.txt file, you can generate with the random walk function in utils.py

Rex