After struggling for a couple of days, I am pretty sure that DynamicGEM cannot perform its temporal graph embedding methods on our dataset for Nov. and Dec. 2010 with almost 125K users.
One approach is filtering users considering some criteria and the other is changing the platform. Since our next step is to test SEERa on the 10-month dataset with a huge amount of users, I decided to change the platform and use PyG-temporal instead of DynamicGEM. PyG-temporal is a huge library and has a lot of functions, specific terms, and training schemes. I have been searching for best-fit methods and settings for SEERa and also testing them on the stream of user graphs. I will report my acquaintance and findings (and hopefully results) here on this issue page.
@hosseinfani
After struggling for a couple of days, I am pretty sure that DynamicGEM cannot perform its temporal graph embedding methods on our dataset for Nov. and Dec. 2010 with almost 125K users. One approach is filtering users considering some criteria and the other is changing the platform. Since our next step is to test SEERa on the 10-month dataset with a huge amount of users, I decided to change the platform and use PyG-temporal instead of DynamicGEM. PyG-temporal is a huge library and has a lot of functions, specific terms, and training schemes. I have been searching for best-fit methods and settings for SEERa and also testing them on the stream of user graphs. I will report my acquaintance and findings (and hopefully results) here on this issue page.