search

shchur / gnn-benchmark

Framework for evaluating Graph Neural Network models on semi-supervised node classification task
https://arxiv.org/abs/1811.05868
MIT License
438 stars 73 forks source link

About Amazon-Computers and Amazon-Photo #12

Open sktsherlock opened 1 year ago

sktsherlock commented 1 year ago

May I ask how to divide the two data sets of Computers and Photo from Amazon-Electronics. I went to the original website(http://jmcauley.ucsd.edu/data/amazon/index_2014.html) to check, but it is still not clear how to divide the two sub-datasets of Computers and Photo. I would be very grateful if you could give me some ideas on what you were doing at that time.

shchur commented 1 year ago

The classes were obtained based on the "categories" field in the metadata file for Electronics. See Sample metadata in the page you linked for an example.

sktsherlock commented 1 year ago

Thank you for your help. I am now trying to figure out how I should determine the label for each node. I see that the description of the node label in PYG is the category to which the node belongs, but I'm not quite sure where I get this category from. For example data with the category ["Electronics", "Camera & Photo", "Video Surveillance", "Surveillance Systems", "Surveillance DVR Kits"], is the third sub-category of it used as the label

shchur commented 1 year ago

If I remember correctly, the categories are nested. That is, Video Surveillance $\subset$ Camera & Photo $\subset$ Electronics. Level-2 labels (Camera & Photo, Computers) were used to generate the two datasets, level-3 labels were used as class labels.

sktsherlock commented 1 year ago

OK ! Thank you for your help.

sktsherlock commented 1 year ago

I would like to ask you in what way you constructed the graph. How do you identify the neighbors of each item? Thank You!

sktsherlock commented 1 year ago

I have now shown its three levels of categories and found that it has 12 categories; I would like to know how you have divided these 12 categories into 8 categories. Thanks! [('Accessories', 9957) ('Lenses', 5267) ('Digital Cameras', 3536) ('Bags & Cases', 3249) ('Tripods & Monopods', 1295) ('Binoculars & Scopes', 1195) ('Lighting & Studio', 647) ('Flashes', 517) ('Film Photography', 460) ('Video', 430) ('Video Surveillance',85)(Printers & Scanners',17)]