Skip to content

Latest commit

 

History

History
9 lines (7 loc) · 622 Bytes

File metadata and controls

9 lines (7 loc) · 622 Bytes

CS150_DGL_CallGraphClassification

Final project for the course CS150-02: Deep Graph Learning @Tufts University, 2023 Spring.

Workflow:

  • Ground truth: Given two call graphs, calculate the directed graph editing distance
  • Experiments: split dataset, train, and test:
    • for 2 input <G1, G2>, loss = dist(embedding1, embedding2), which is the predicted distance value generated by the GNN model, taking the two graphs as input and producing two embedding vectors as intermediate values accordingly.
    • train multiple times and fine-tune with various dataset, to obtain an optimal model with an accuracy score