search

cran2367 / sgt

Sequence Graph Transform
107 stars 21 forks source link

Creating network graph from SGT Embedding #13

Open sdattaIDBS opened 1 year ago

sdattaIDBS commented 1 year ago

Hello,

Thank you for developing and sharing a unique algorithm for sequence embedding. I am currently implementing your SGT algorithm for protein embedding. While reading your paper, I could see in fig 2(a) that SGT embedding could be transformed to network graph. I intend to create graph and feed to Vanilla Graph Neural Network (am new to GNN). I am facing issue in creating graph out of the SGT embedding. I was thinking to unwrap the tuple columns to be as source and destination nodes and the decimal values to be as weight values. However. I was struggling with the graph.x (torch geometric) dimension.

I was wondering if you could help me on how to create create network graph from SGT embedding.

Thanks in advance.

cran2367 commented 1 year ago

Hi, SGT provides the adjacency matrix. I believe you are already using the adjacency matrix to create the (source, destination, weight) tuples. I am now sure of how to work with graph.x. If there is anything specific you are looking in SGT, please let me know.