One-step corrected projected stochastic gradient descent for statistical estimation

arXiv:2306.05896v2 Announce Type: replace-cross
Abstract: A generic, fast and asymptotically efficient method for parametric estimation is described. It is based on the projected stochastic gradient descent on the log-likelihood function corrected by a single step of the Fisher scoring algorithm. We show theoretically and by simulations that it is an interesting alternative to the usual stochastic gradient descent with averaging or the adaptative stochastic gradient descent.

abstract algorithm arxiv fisher function gradient likelihood math.st parametric scoring show simulations statistical stat.ml stat.th stochastic type