Filtering, State Estimation, and Other Forms of Signal Processing

01 Mar 2015 22:10

Filtering. Is the Kalman filter like the EM algorithm, in any meaningful sense? (Yes; where'd that paper by Gharamani showing this go?) What is the Wonham filter, exactly? What are the estimation properties of the parameters in these filters? How bad is it to use the standard linear filters (Wiener, Kalman) on nonlinear systems? What do the existing nonlinear filters look like? (Hidden Markov models are one class of nonlinear filter; they have various drawbacks, mostly about needing to choose the architecture a priori, and it being hard to tell if you're using the wrong architecture, or the process is just intrinsically ugly.)

Nonlinear filtering.

Independent component analysis.

See also: Control Theory; Information Geometry; Monte Carlo and Stochastic Simulation; Time Series