Closed ArtPoon closed 6 years ago
In brief, our implementation (and others) are consistently giving half the distance of the Kuhner-Yamato implementation.
Making this low priority, let's not expend too much resources into diagnosing someone else's implementation of RF distance
We did not actually implement RF.dist, KF.dist, and path.dist metrics. We used the package phangorn
for this. Phangorn requires the igraph library to be re-compiled when bringing in this package (even though Kaphi itself has already compiled igraph). Would it be better simply to re-implement these distances again ourselves?
For now use phangorn
. Let's see how useful these distances will be before we think about re-implementing anything.
See issue #85