I ran node2vec on the directed graph cora. But, this classification micro f1 score (~ 0.35) was worse than LINE-1st (~ 0.80).
I guess this reason is the definition of walksVV. If random walker reaches a node that has 0 out degree, alias sampler will stop walking, but walksVV will contain many 0.
I ran node2vec on the directed graph cora. But, this classification micro f1 score (~ 0.35) was worse than LINE-1st (~ 0.80).
I guess this reason is the definition of
walksVV
. If random walker reaches a node that has 0 out degree, alias sampler will stop walking, butwalksVV
will contain many0
../node2vec -i:cora_edgeslist.txt -o:cora_walk -d:100 -p:1 -q:1 -dr -v -ow