fani-lab / SEERa

A framework to predict the future user communities in a text streaming social network based on the users’ topics of interest.
Other
4 stars 5 forks source link

SEERa refactoring - Moving toward PyG Temporal #77

Open soroush-ziaeinejad opened 1 year ago

soroush-ziaeinejad commented 1 year ago

In this issue page, we will report the challenges through refactoring steps of SEERa and mostly for the graph embedding layer. Multiple steps can be defined for this task as listed below (the list is going to be updated regularly):

  1. adding (followship only/ latent only/ followship and latent) options
  2. extracting followship information out of the data
  3. adding topics as node features (if the latent option is on)
  4. creating edges as node connections (followship or latent based on the user choice)
  5. predict node features to obtain user interest in the future
  6. predict edges to obtain users' connections in the future
  7. preparing the dataset to be fed to pyg temporal nn models
  8. removing dynamicgem and using pyg temporal for graph embedding
  9. check if sentence transformers can embed raw strings better than simple BOW https://www.sbert.net/
  10. Which loss functions are suitable for which problems?