| Paper: | MLSP-P1.6 | ||
| Session: | Blind Source Separation and ICA | ||
| Time: | Tuesday, May 18, 15:30 - 17:30 | ||
| Presentation: | Poster | ||
| Topic: | Machine Learning for Signal Processing: Blind Signal Separation and Independent Component Analysis | ||
| Title: | A QUASI-OPTIMALLY EFFICIENT ALGORITHM FOR INDEPENDENT COMPONENT ANALYSIS | ||
| Authors: | John Weng; Michigan State University | ||
| Nan Zhang; Michigan State University | |||
| Abstract: | We propose an incremental algorithm for independent componentanalysis (ICA), that is guided by the statistical efficiency.Starting from a $\ell ^ \infty$ norm sparseness measure contrastfunction, we derive the learning algorithm based on awinner-take-all learning mechanism. It avoids the optimization ofhigh order non-linear function or density estimation, which havebeen used by other ICA methods, such as negentropy approximation,infomax, and maximum likelihood estimation based methods. We showthat when the latent independent random variables aresuper-Gaussian distributions, the network efficiently extracts the independent components. We observed a much faster convergence than other ICA methods. | ||
| Back | |||
Home -||-
Organizing Committee -||-
Technical Committee -||-
Technical Program -||-
Plenaries
Paper Submission -||-
Special Sessions -||-
ITT -||-
Paper Review -||-
Exhibits -||-
Tutorials
Information -||-
Registration -||-
Travel Insurance -||-
Housing -||-
Workshops