This algorithm generalizes the JADE algorithm, as it provides JADE when k is set to the number of dimensions. Otherwise k can be considered as a way to reduce the number of cumulant matrices to be jointly diagonalized. Hence small values of k speed up the method considerably in high-dimensional cases. In general, k can be considered as maximum number of underlying identical sources.
The function uses FOBI as an initial estimate and frjd for the joint diagonalization.
A list with class 'bss' containing the following components:
A
The estimated mixing matrix.
W
The estimated unmixing matrix.
S
Matrix with the estimated independent components.
Xmu
The location of the original data.
Arguments
X
Numeric data matrix or dataframe.
k
integer value between 1 and the number of columns of X. Default is 1.
eps
Convergence tolerance.
maxiter
Maximum number of iterations.
na.action
A function which indicates what should happen when the data
contain 'NA's. Default is to fail.
Author
Jari Miettinen
Details
The order of the estimated components is fixed so that their fourth moments are in the decreasing order.
References
Miettinen, J., Nordhausen, K., Oja, H. and Taskinen, S. (2013), Fast Equivariant JADE,
In the Proceedings of 38th IEEE International Conference on Acoustics, Speech, and Signal Processing (ICASSP 2013), 6153--6157.
Miettinen, J., Nordhausen, K. and Taskinen, S. (2017), Blind Source Separation Based on Joint Diagonalization in R: The Packages JADE and BSSasymp, Journal of Statistical Software, 76, 1--31, <doi:10.18637/jss.v076.i02>.