search

uoguelph-mlrg / LDG

PyTorch code for "Learning Temporal Attention in Dynamic Graphs with Bilinear Interactions"
https://arxiv.org/abs/1909.10367
Other
98 stars 24 forks source link

Pre-processed SocialEvolution dataset? #3

Closed zhh0998 closed 4 years ago

zhh0998 commented 4 years ago

Will the preprocessed version of the SocialEvolution dataset be publicly available？Why the GITHUB dataset provided is different from SocialEvolution？

bknyaz commented 4 years ago

The SocialEvolution and Github datasets are two different datasets available in respective folders. The original Github dataset is huge, so we provide a preprocessed version. SocialEvolution is small and a preprocessed version is prepared each time the training script is executed. All events for train/test splits of both datasets are stored in the all_events variable: https://github.com/uoguelph-mlrg/LDG/blob/master/social_data_loader.py#L101

all_events is a list of tuples: (node1, node2, type of event, timestamp)

So if you want to try other methods on these datasets, I would suggest saving all_events to .pkl file or something like that and loading them in your framework.