Description Usage Arguments Details Value
Provides a simple function to bluid and fit a block mixture of von Mises-Fisher for data co-clustering.
1 |
X |
A sparse or dense data matrix, where rows denote objets lying on a unit-hypersphere and column denote features. Supported format include Matrix, dgCMatrix, dgTMatrix, etc. |
k |
The number of mixture compoenents (or clusters). |
control |
A list if additional parameters, see Details. |
... |
A list of additional parameters (overriding those specified in control). |
In addition to the parameters described above, the function dbmovMF supports other useful parameters including:
max_iter: the maximum number of iteration
max_st_iter: the maximum number of stochatic iterations. Only useful when performing learning using either SAEMb or CAEMb
n_init: number of time the algorithm will be run with different initializations. The final results correspond to the bust run in terms of log likelihood.
tol: tolerance relatve to the log likelihood to declar convergence.
equal_prop: if true then then all mixture proportion will be consider to be equal, and the mixture propotions parmameters alpha will be ignored.
fit_algo: learning algorithm to use. Supported algorithms include: EMb, CEMb, SEMb, SAEMb and CAEMb.
equal_kappa: if true then all concentration parameters kappa of the differrent mixture components will be equal
kappa_: The value of the concentration parameter. This parameter is only effective when euqal_kappa = TRUE.
row_init: a partition of rows into k cluster to be used for initialization. Could be a vector if n_init = 1 of a matrix of size (n_init x n_rows) if n_init>1. Default = NULL.
col_init: same as row_init but for columns.
returns an object of class dbmovMF including the following attributes:
rowcluster: the partition of rows into k clusters.
colcluster: the partition of columns into k clusters.
kappa_: a vector of concentration parameters.
alpha: a vector of cluster proportions.
alpha: a vector of cluster proportions.
ll: the values of the complete data log likelihood over iterations.
Iter: the total number of iteration needed for convergence.
Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.