Hi
I am looking for parallel implementation of hierarchical clustering, the
equivalent to "hclust" in the "fpc" package.
I found "hcluster" from "amap" package:
hcluster(x, method = "euclidean", diag = FALSE, upper = FALSE,
link = "complete", members = NULL, nbproc = 2,
doubleprecision = TRUE)
It takes a data matrix, computes distance matrix then do clustering.
However in my application, /i have to compute the distance matrix and
use it later anyway. So hcluster is re-computing the distance which is a
waste of time, as my data is very large scale.
Is there anyway hcluster could just use a pre-computed distance object,
or obtain the distance object from hcluster, so I can avoid
double-computing the distane object?
Or more general question is, if there is a parallel implementation of
hierarchical clustering that takes input a distance matrix, rather than
the raw data matrix?
Many thanks!
---
This email has been checked for viruses by Avast antivirus software.