A Python implementation of the ZORO algorithm, as introduced in Zeroth-order regularized optimization (ZORO): Approximately sparse gradients and adaptive sampling by Cai, McKenzie, Yin and Zhang. This paper is accepted to SIAM Journal on Optimization (SIOPT). Preprint available here .
ZORO, and its adaptive variant AdaZORO, are implemented as classes in optimizers.py. We use the BaseOptimizer class as in this Repo.
Python 3.5+. For proximal operators the pyproximal package is required.
See Test.py, Test_Prox.py and Test_Ada.py for examples of using ZORO.
Feel free to contact us at mckenzie@math.ucla.edu or hqcai@ucf.edu.
If you find this code useful please cite the following work:
HanQin Cai, Daniel Mckenzie, Wotao Yin, and Zhenliang Zhang. Zeroth-Order Regularized Optimization (ZORO): Approximately Sparse Gradients and Adaptive Sampling. arXiv preprint arXiv: 2003.13001.
Bibtex:
@article{cai2020zeroth,
title={Zeroth-order regularized optimization (zoro): Approximately sparse gradients and adaptive sampling},
author={Cai, HanQin and Mckenzie, Daniel and Yin, Wotao and Zhang, Zhenliang},
journal={arXiv preprint arXiv:2003.13001},
year={2020}
}