GATNE
GATNE copied to clipboard
THUDM
Source code and dataset for KDD 2019 paper "Representation Learning for Attributed Multiplex Heterogeneous Network"
My dataset has 6 types of edge and 3 types of nodes. When I run this model with main.py without schema, it shows the failure as follows. Traceback (most recent...
Where are the trainable weights `W^{k}` as well as the activation in equation (2)? I can't find it in the code. Are these weights be missed?
Hello, how do you get candidate nodes when using GATNE for link prediction?
In function `generate_neighbors`, there is a process to lookup encode node id from `vocab` for all nodes in the graph. The `vocab` is generated from all random walks. However, there...
is:issue is:open Whether GATNE can handle nodes with different feature dimensions？
{'epoch': 0, 'iter': 0, 'avg_loss': 19.13713836669922, 'loss': 19.137138} {'epoch': 0, 'iter': 5000, 'avg_loss': 9.393543103258029, 'loss': 5.517972} epoch 0: 100%|| 7767/7767 [05:38
你好： 两类节点，user和item节点都添加到里面，我是想着把user节点和item的节点分开算auc，例如评价user节点，两个user消费同一个item算正样本，然后负样本随机采样；评价item节点，两个item被同一个user消费算正样本，然后负样本随机采样。你觉得这样评价可行吗？
vocab是通过all_walks生成的，但是generate_neighbors是遍历network_data的。如果generate_walks过程中某一个节点不幸未被选中参与到walks中，那么在generate_neighbors阶段就会KeyError
