One-step corrected projected stochastic gradient descent for statistical estimation
Statistics Theory
2024-04-16 v2 Machine Learning
Statistics Theory
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.
Cite
@article{arxiv.2306.05896,
title = {One-step corrected projected stochastic gradient descent for statistical estimation},
author = {Alexandre Brouste and Youssef Esstafa},
journal= {arXiv preprint arXiv:2306.05896},
year = {2024}
}