Adapting to Nonstationary Behavior. Examples from Geophysics and Cosmology
Thursday, April 25, 2002 - 3:00pm - 4:00pm
Bias in inverse problem estimates can be reduced by selecting models that can adapt to the true nature of the unknown signals. We present some examples of applications of mixture models that can adapt to nonstationary behavior in deconvolution problems in geophysics and in point source modeling problems in cosmology. The deconvolution method generalizes the classical Wiener-Levinson approach while the point source application can be used to model coupled homogeneous random fields with prescribed cross-spectrum and marginal structure.