A generic, fast and asymptotically efficient method for parametric estimation is described. It is based on the stochastic gradient descent on the loglikelihood function corrected by a single step of the Fisher scoring algorithm. We show theoretically and by simulations in the i.i.d. setting that it is an interesting alternative to the usual stochastic gradient descent with averaging or the adaptative stochastic gradient descent.

翻译：本文描述了一种通用、快速且渐近有效的参数估计方法。该方法基于对数似然函数的随机梯度下降，并通过Fisher得分算法的一步校正完成。我们在独立同分布设定下通过理论推导和仿真实验证明，该方法相较于常规带平均的随机梯度下降或自适应随机梯度下降，是一种具有竞争力的替代方案。