Description Usage Arguments Details Value References See Also Examples
An parallelised implementation of a Jarník (Prim/Dijkstra)like algorithm for determining a(*) minimum spanning tree (MST) of a complete undirected graph representing a set of n points with weights given by a pairwise distance matrix.
(*) Note that there might be multiple minimum trees spanning a given graph.
1 2 3 4 5 6 7 8 9 10 11 12 13 14 
d 
either a numeric matrix (or an object coercible to one,
e.g., a data frame with numericlike columns) or an
object of class 
... 
further arguments passed to or from other methods. 
distance 
metric used to compute the linkage, one of:

M 
smoothing factor; 
cast_float32 
logical; whether to compute the distances using 32bit instead of 64bit precision floatingpoint arithmetic (up to 2x faster). 
verbose 
logical; whether to print diagnostic messages and progress information. 
If d
is a numeric matrix of size n p,
the n (n1)/2 distances are computed on the fly, so that O(n M)
memory is used.
The algorithm is parallelised; set the OMP_NUM_THREADS
environment
variable Sys.setenv
to control the number of threads
used.
Time complexity is O(n^2) for the method accepting an object of
class dist
and O(p n^2) otherwise.
If M
>= 2, then the mutual reachability distance m(i,j) with smoothing
factor M
(see Campello et al. 2015)
is used instead of the chosen "raw" distance d(i,j).
It holds m(i, j)=\max(d(i,j), c(i), c(j)), where c(i) is
d(i, k) with k being the (M
1)th nearest neighbour of i.
This makes "noise" and "boundary" points being "pulled away" from each other.
Genie++ clustering algorithm (see gclust
)
with respect to the mutual reachability distance gains the ability to
identify some observations are noise points.
Note that the case M
= 2 corresponds to the original distance, but we are
determining the 1nearest neighbours separately as well, which is a bit
suboptimal; you can file a feature request if this makes your data analysis
tasks too slow.
Matrix of class mst
with n1 rows and 3 columns:
from
, to
and dist
. It holds from
< to
.
Moreover, dist
is sorted nondecreasingly.
The ith row gives the ith edge of the MST.
(from[i], to[i])
defines the vertices (in 1,...,n)
and dist[i]
gives the weight, i.e., the
distance between the corresponding points.
The method
attribute gives the name of the distance used.
The Labels
attribute gives the labels of all the input points.
If M
> 1, the nn
attribute gives the indices of the M
1
nearest neighbours of each point.
Jarník V., O jistém problému minimálním, Práce Moravské Přírodovědecké Společnosti 6 (1930) 57–63.
Olson C.F., Parallel algorithms for hierarchical clustering, Parallel Comput. 21 (1995) 1313–1325.
Prim R., Shortest connection networks and some generalisations, Bell Syst. Tech. J. 36 (1957) 1389–1401.
Campello R., Moulavi D., Zimek A., Sander J., Hierarchical density estimates for data clustering, visualization, and outlier detection, ACM Transactions on Knowledge Discovery from Data 10(1) (2015) 5:1–5:51.
emst_mlpack()
for a very fast alternative
in case of (very) lowdimensional Euclidean spaces (and M
= 1).
1 2 3 4 
Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.