How can i get the report result of 0.828 in PPI dataset?

[holoword]

I run the PPI dataset in the default configuration, but what I only get the best result is 0.752, how can I get the paper result which is 0.828.

[urielsinger]

Hey, This is a beta version of the code. Working on pushing a final version.

[NameIsAlreadyTakenY]

Using the default parameters, I get the following result on PPI dataset:

tnodeembed: {'auc': 0.774697936929617, 'f1_micro': 1.0, 'f1_macro': 1.0} node2vec: {'auc': 0.7457705787129938, 'f1_micro': 1.0, 'f1_macro': 1.0}

Interesting results!

[DarkPar]

set the trainskip to 1 'train_skip': 1, # down sample the training set then you can get a better result i got tnodeembed: {'auc': 0.809940061974782, 'f1_micro': 0.6632322490147619, 'f1_macro': 0.6310469252486426} node2vec: {'auc': 0.7658663076403662, 'f1_micro': 0.6160577115757131, 'f1_macro': 0.5585983049608252}

lower f1 score, but higher auc