Skip to main content
. 2022 Aug 17;18(9):853–864. doi: 10.1039/d1mo00521a
Unsupervised (TripleWalk and SGNS)
Parameter Value
Batch size 128
Learning rate 0.0004
Embedding dimension 256
Random walk length 17
Random walks per node 6
Early stopping delta (loss) 0.1
Early stopping patience (epochs) 5
Number of negative samples per positive sample 2.0