One-step corrected projected stochastic gradient descent for statistical estimation

Alexandre Brouste, Youssef Esstafa

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…