Uniqueness and stability for the solution of a nonlinear least squares problem

Author:

Huang Meng,Xu Zhiqiang

Abstract

In this paper, we focus on the nonlinear least squares problem: min x H d | A x | b \min _{{\boldsymbol {x}}\in \mathbb {H}^d}\|\lvert A{\boldsymbol {x}}\rvert -{\boldsymbol {b}}\| where A H m × d A\in \mathbb {H}^{m\times d} , b R m {\boldsymbol {b}}\in \mathbb {R}^m with H { R , C } \mathbb {H}\in \left \{\mathbb {R},\mathbb {C}\right \} and consider the uniqueness and stability of solutions. This problem arises in applications such as phase retrieval and absolute value rectification neural networks. While several results have been developed to characterize the uniqueness and stability of solutions when b = | A x 0 | {\boldsymbol {b}}=\lvert A{\boldsymbol {x}}_0\rvert for some x 0 H d {\boldsymbol {x}}_0\in \mathbb {H}^d , no existing results address the case where b {\boldsymbol {b}} is arbitrary. In this paper, we investigate the uniqueness and stability of solutions for the more general case where b {\boldsymbol {b}} is not necessarily equal to | A x 0 | \lvert A{\boldsymbol {x}}_0\rvert for any x 0 H d {\boldsymbol {x}}_0\in \mathbb {H}^d . We prove that for any matrix A H m × d A\in \mathbb {H}^{m\times d} , there is always a vector b R m {\boldsymbol {b}}\in \mathbb {R}^m for which the solution to the nonlinear least squares problem is not unique. However, we show that such “bad” vectors b {\boldsymbol {b}} are negligible in practice; specifically, if b R m {\boldsymbol {b}}\in \mathbb {R}_{ }^m does not lie in some measure zero set, then the solution is unique. Furthermore, we establish certain conditions under which the solution is guaranteed to be unique. Regarding the stability of solutions, we prove that the solution is not uniformly stable. However, if we restrict the vectors b {\boldsymbol {b}} to a convex set where the solution to the least squares problem is unique, then the solution becomes stable. To the best of our knowledge, our results represent the first theoretical results of the uniqueness and stability of solutions for the nonlinear least squares problem.

Funder

National Social Science Fund of China

Publisher

American Mathematical Society (AMS)

Subject

Applied Mathematics,Computational Mathematics,Algebra and Number Theory

Cited by 1 articles. 订阅此论文施引文献 订阅此论文施引文献,注册后可以免费订阅5篇论文的施引文献,订阅后可以查看论文全部施引文献

1. The performance of the amplitude-based model for complex phase retrieval;Information and Inference: A Journal of the IMA;2024-01-01

同舟云学术

1.学者识别学者识别

2.学术分析学术分析

3.人才评估人才评估

"同舟云学术"是以全球学者为主线,采集、加工和组织学术论文而形成的新型学术文献查询和分析系统,可以对全球学者进行文献检索和人才价值评估。用户可以通过关注某些学科领域的顶尖人物而持续追踪该领域的学科进展和研究前沿。经过近期的数据扩容,当前同舟云学术共收录了国内外主流学术期刊6万余种,收集的期刊论文及会议论文总量共计约1.5亿篇,并以每天添加12000余篇中外论文的速度递增。我们也可以为用户提供个性化、定制化的学者数据。欢迎来电咨询!咨询电话:010-8811{复制后删除}0370

www.globalauthorid.com

TOP

Copyright © 2019-2024 北京同舟云网络信息技术有限公司
京公网安备11010802033243号  京ICP备18003416号-3