streaming_node2vec

This repository is a part of the course CS430 Algorithms for Data Science at IIT Gandhinagar. We have used three different sampling algorithms at ~10% and ~1% and generated node2vec feature vectors using that. Then we have used Support Vector Machines for Link Predictions. The presentation contains entire training procedure.

Sampling Techniques

• Reservoir Sampling

• Graph Priority Sampling

  • Using number of triangles completed by the edge as weight
  • Minimum degree of the nodes of the upcoming edge as weight

• Graph Sample and Hold

  ​

  Link Prediction

  Supervised Link Prediction