Complexity Penalized M-Estimation

摘要: We present very fast algorithms for the exact computation of estimators time series, based on complexity penalized log-likelihood or M-functions. The apply to a wide range functionals with morphological constraints, in particular Potts Blake–Zisserman functionals. latter are discrete versions celebrated Mumford–Shah All such contain model parameters. Our allow optimization not only each separate parameter, but even all parameters simultaneously. This allows examination models sense family approach. accompanied by series illustrative examples from molecular biology.