Yuriy Tyshetskiy

Directed HinSAGE

1

### Description Given that we have an implementation of directed SampledBFS walks, we should be able to implement HinSAGE for directed graphs. See #465 for implementation of directed GraphSAGE (This...

Loading a saved HinSAGE link prediction model fails

12

### Description @adocherty this might happen with the yelp example as well! (once the `test()` function is implemented) In branch `issue#186`: When loading a HinSAGE+link regression model, saved by .train()...

Before the fix, saving the model failed when `pretrain` was set to `False`, with the following error: ``` File "/Deep-SAD-PyTorch/src/DeepSAD.py", line 133, in save_model ae_net_dict = self.ae_net.state_dict() if save_ae else...

Stochastic predictions with GraphSAGE

1

During training, each graphsage layer takes uniform samples from each node's neighbours, a new sample at each training iteration. With a large number of samples taken during the training, the...