Newton method for the ICA mixture model

We derive an asymptotic Newton algorithm for Quasi Maximum Likelihood estimation of the ICA mixture model, using the ordinary gradient and Hessian. The probabilistic mixture framework can accommodate non-stationary environments and arbitrary source densities. We prove asymptotic stability when the source models match the true sources. An application to EEG segmentation is given.

• Type: Conference
• Year: 2008
• Where: ICASSP
• Authors: Jason A. Palmer, Scott Makeig, Kenneth Kreutz-Delgado, Bhaskar D. Rao
• Keywords: ICA Mixture Model, ICASSP 2008, Probabilistic Mixture Framework, Quasi Maximum Likelihood, Signal Processing