search

snap-stanford / neural-subgraph-learning-GNN

340 stars 64 forks source link

Ratio of Train set and Test set size #4

Open abhishekrajgaria opened 3 years ago

abhishekrajgaria commented 3 years ago

As written in common/data.py in load_dataset the ratio of train_set : test_set is 80 : 20,

but as we randomly generate positive and negative query-target pair, (balance case) we are getting 4096 graphs, for both training set and testing set, (imbalance case) we are getting 2048 graphs for both training set and testing set,

So won't that be an issue as the size of train_set is same as test_set?

qema commented 3 years ago

Can't confirm right now but I believe there is an option --val_size to change the validation set size. Agree val_size should be increased during a rigorous evaluation

On Oct 27, 2020, at 05:23, Abhishek Rajgaria notifications@github.com wrote:

 As written in common/data.py in load_dataset the ratio of train_set : test_set is 80 : 20,

but as we randomly generate positive and negative query-target pair, (balance case) we are getting 4096 graphs, for both training set and testing set, (imbalance case) we are getting 2048 graphs for both training set and testing set,

So won't that be an issue as the size of train_set is same as test_set?

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub, or unsubscribe.