On use of the EM algorithm for penalized likelihood estimation

Abstract

The EM algorithmis a popular approach to maximuml ikelihoode stimationb ut has not been muchu sed forp enalizedl ikelihoodo r maximuma posteriori estimation.T his paper discussesp ropertieos f theE M algorithmin suchc ontextsc, oncentratinogn rateso f convergence, and presentsa n alternativet hati s usuallym ore practicala nd convergesa t least as quickly