Open carissableker opened 2 months ago
Hi,
This permutation is there only because of the way I evaluate the predicted ranking of edges. When I know the true network, I typically evaluate the predicted ranking by computing the area under the precision-recall curve (AUPR). Given the way I compute this AUPR in practice, I would get an AUPR that is over-estimated if I throw away the edges with score 0.
What is the reasoning behind randomly permuting the edges with score 0 in
get_link_list
?https://github.com/vahuynh/dynGENIE3/blob/ce8a35af949afc3ae69e38541a966379ae4a9361/dynGENIE3_python/dynGENIE3.py#L111
Does it make sense to simply throw these edges away?