This is an unsupervised model for node classification of heterogeneous graph. I changed the framework from 0.4.3.post2 to 0.6.1 and found that the results of the model are completely different, even only half. In 0.6.1, there will be a sudden increase in the loss function. What is the reason for this?
Several questions:
- What is the unsupervised model you were running?
- Did you make any changes across versions?
- Could you post the code or provide a link to it?
- How large is the performance gap?
When training the model, the loss is normal at the beginning, but then there will be a sudden increase in the loss, with the following warning.
/root/yes/envs/SPGCN/lib/python3.6/site-packages/sklearn/linear_model/_logistic.py:765: ConvergenceWarning: lbfgs failed to converge (status=1):
STOP: TOTAL NO. of ITERATIONS REACHED LIMIT.
But I didn’t find this function in my project code.
This is the model I use, without any modification to the code.
This topic was automatically closed 30 days after the last reply. New replies are no longer allowed.