nmfSeed lists and retrieves NMF seeding methods.
getNMFSeed is an alias for nmfSeed.
existsNMFSeed tells if a given seeding method
exists in the registry.
nmfSeed(name = NULL, ...) getNMFSeed(name = NULL, ...)
existsNMFSeed(name, exact = TRUE)
access key of a seeding method stored in
registry. If missing, nmfSeed returns the list of
all available seeding methods.
extra arguments used for internal calls
a logical that indicates if the access key should be matched exactly or partially.
Currently the internal registry contains the following
seeding methods, which may be specified to the function
nmf via its argument seed using
their access keys:
The entries of each factors are drawn from a uniform distribution over \([0, max(x)]\), where $x$ is the target matrix.
Nonnegative Double Singular Value Decomposition.
The basic algorithm contains no randomization and is based on two SVD processes, one approximating the data matrix, the other approximating positive sections of the resulting partial SVD factors utilising an algebraic property of unit rank matrices.
It is well suited to initialise NMF algorithms with sparse factors. Simple practical variants of the algorithm allows to generate dense factors.
Reference: Boutsidis et al. (2008)
Uses the result of an Independent Component
Analysis (ICA) (from the fastICA package). Only
the positive part of the result are used to initialise
the factors.
Fixed seed.
This method allows the user to manually provide initial values for both matrix factors.
Boutsidis C and Gallopoulos E (2008). "SVD based initialization: A head start for nonnegative matrix factorization." _Pattern Recognition_, *41*(4), pp. 1350-1362. ISSN 00313203, <URL: http://dx.doi.org/10.1016/j.patcog.2007.09.010>, <URL: http://linkinghub.elsevier.com/retrieve/pii/S0031320307004359>.
# roxygen generated flag
options(R_CHECK_RUNNING_EXAMPLES_=TRUE)
# list all registered seeding methods
nmfSeed()
# retrieve one of the methods
nmfSeed('ica')
Run the code above in your browser using DataLab