Can't do node_classification tasks on panther datasets
wangzeyu135798 opened this issue · comments
zeyuwang commented
When I use cikm dataset in panther directory do downstream node_classification task, in data_utils class SSSingleDataset, self.data = Data(x=None, edge_index = egde_index,y = None), while in graph_dataset.py line 434 self.num_classes = self.data.y.shape[1], this self.data is Data object, self.data.y doesn't exist so it runs wrong.
Qibin Chen commented
Hi @wangzeyu135798 ,
Datasets in the panther directory are for the Similarity Search task. They don't have node labels and don't support the node classification task. Please check out the experimental sections in the paper for details.