TY - JOUR
T1 - Complex infomax
T2 - Convergence and approximation of infomax with complex nonlinearities
AU - Calhoun, Vince
AU - Adali, Tülay
N1 - Funding Information:
This work was supported in part by the National Science Foundation Career Award, NSF NCR-9703161 (to TA) and the National Institutes of Health 1 R01 EB 000840-01 (to VC).
PY - 2006/8
Y1 - 2006/8
N2 - Independent component analysis (ICA) for separating complex-valued sources is needed for convolutive source-separation in the frequency domain, or for performing source separation on complex-valued data, such as functional magnetic resonance imaging or radar data. Previous complex Infomax approaches that use nonlinear functions in the updates have proposed using bounded (and hence non-analytic) nonlinearities. In this paper, we propose using an analytic (and hence unbounded) complex nonlinearity for Infomax for processing complex-valued sources. We show by simulation examples that using an analytic nonlinearity for processing complex data has a number of advantages. First, when compared to split-complex approaches (i.e., approaches that split the real and imaginary data into separate channels), the shape of the performance surface is improved resulting in better convergence characteristics. We also show that using an analytic complex-valued function for the nonlinearity is more effective in generating the higher order statistics required to establish independence when compared to complex nonlinear functions, i.e., functions that are ℂ → ℂ.
AB - Independent component analysis (ICA) for separating complex-valued sources is needed for convolutive source-separation in the frequency domain, or for performing source separation on complex-valued data, such as functional magnetic resonance imaging or radar data. Previous complex Infomax approaches that use nonlinear functions in the updates have proposed using bounded (and hence non-analytic) nonlinearities. In this paper, we propose using an analytic (and hence unbounded) complex nonlinearity for Infomax for processing complex-valued sources. We show by simulation examples that using an analytic nonlinearity for processing complex data has a number of advantages. First, when compared to split-complex approaches (i.e., approaches that split the real and imaginary data into separate channels), the shape of the performance surface is improved resulting in better convergence characteristics. We also show that using an analytic complex-valued function for the nonlinearity is more effective in generating the higher order statistics required to establish independence when compared to complex nonlinear functions, i.e., functions that are ℂ → ℂ.
KW - ICA
KW - Infomax
KW - fMRI
KW - independent component analysis
UR - http://www.scopus.com/inward/record.url?scp=33746386881&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=33746386881&partnerID=8YFLogxK
U2 - 10.1007/s11265-006-7514-5
DO - 10.1007/s11265-006-7514-5
M3 - Article
AN - SCOPUS:33746386881
SN - 1387-5485
VL - 44
SP - 173
EP - 190
JO - Journal of VLSI Signal Processing Systems for Signal, Image, and Video Technology
JF - Journal of VLSI Signal Processing Systems for Signal, Image, and Video Technology
IS - 1-2
ER -