Skip to content

Latest commit

 

History

History
10 lines (7 loc) · 472 Bytes

README.md

File metadata and controls

10 lines (7 loc) · 472 Bytes

heteSAGE

heteSAGE is semi-supervised graphSAGE model that support metapaths of heterogeneous features.

  1. Heterogeneous features include sparse and dense features of multi type of vertexes.
  2. Share weights between multi-metapaths
  3. Use the attention mechanism to aggregate the results of multi-metapaths embedding of the same node type
  4. Use graphSAGE aggregate method to aggregate multi-hops features

trainning sample contains <target id, context id, ..., label>