@inproceedings{zhang2019a,
author = {Zhang, Junyu and Xiao, Lin},
title = {A Stochastic Composite Gradient Method with Incremental Variance Reduction},
booktitle = {Neural Information Processing Systems},
year = {2019},
month = {December},
abstract = {We consider the problem of minimizing the composition of a smooth (nonconvex) function and a smooth vector mapping, where the inner mapping is in the form of an expectation over some random variable or a finite sum. We propose a stochastic composite gradient method that employs incremental variance-reduced estimators for both the inner vector mapping and its Jacobian. We show that this method achieves the same orders of complexity as the best known first-order methods for minimizing expected-value and finite-sum nonconvex functions, despite the additional outer composition which renders the composite gradient estimator biased. This finding enables a much broader range of applications in machine learning to benefit from the low complexity of incremental variance-reduction methods.},
url = {http://approjects.co.za/?big=en-us/research/publication/a-stochastic-composite-gradient-method-with-incremental-variance-reduction-2/},
}