TY - JOUR

T1 - Application of gradient descent algorithms based on geodesic distances

AU - Duan, Xiaomin

AU - Sun, Huafei

AU - Peng, Linyu

N1 - Funding Information:
Xiaomin DUAN was supported by National Science and Technology Major Project of China (Grant No. 2016YFF02030012), National Natural Science Foundation of China (Grant No. 61401058) and Natural Science Foundation of Liaoning Province (Grant No. 20180550112). Huafei SUN was partially supported by National Natural Science Foundation of China (Grant No. 61179031). Linyu PENG was supported by JSPS Grant-in-Aid for Scientific Research (Grant No. 16KT0024), the MEXT “Top Global University Project”, Waseda University Grant for Special Research Projects (Grant Nos. 2019C–179, 2019E–036), and Waseda University Grant Program for Promotion of International Joint Research.
Publisher Copyright:
© 2020, Science China Press and Springer-Verlag GmbH Germany, part of Springer Nature.

PY - 2020/5/1

Y1 - 2020/5/1

N2 - In this paper, the Riemannian gradient algorithm and the natural gradient algorithm are applied to solve descent direction problems on the manifold of positive definite Hermitian matrices, where the geodesic distance is considered as the objective function. The first proposed problem is the control for positive definite Hermitian matrix systems whose outputs only depend on their inputs. The geodesic distance is adopted as the difference of the output matrix and the target matrix. The controller to adjust the input is obtained such that the output matrix is as close as possible to the target matrix. We show the trajectory of the control input on the manifold using the Riemannian gradient algorithm. The second application is to compute the Karcher mean of a finite set of given Toeplitz positive definite Hermitian matrices, which is defined as the minimizer of the sum of geodesic distances. To obtain more efficient iterative algorithm than traditional ones, a natural gradient algorithm is proposed to compute the Karcher mean. Illustrative simulations are provided to show the computational behavior of the proposed algorithms.

AB - In this paper, the Riemannian gradient algorithm and the natural gradient algorithm are applied to solve descent direction problems on the manifold of positive definite Hermitian matrices, where the geodesic distance is considered as the objective function. The first proposed problem is the control for positive definite Hermitian matrix systems whose outputs only depend on their inputs. The geodesic distance is adopted as the difference of the output matrix and the target matrix. The controller to adjust the input is obtained such that the output matrix is as close as possible to the target matrix. We show the trajectory of the control input on the manifold using the Riemannian gradient algorithm. The second application is to compute the Karcher mean of a finite set of given Toeplitz positive definite Hermitian matrices, which is defined as the minimizer of the sum of geodesic distances. To obtain more efficient iterative algorithm than traditional ones, a natural gradient algorithm is proposed to compute the Karcher mean. Illustrative simulations are provided to show the computational behavior of the proposed algorithms.

KW - Karcher mean

KW - Riemannian gradient algorithm

KW - Toeplitz positive definite Hermitian matrix

KW - natural gradient algorithm

KW - system control

UR - http://www.scopus.com/inward/record.url?scp=85083079843&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85083079843&partnerID=8YFLogxK

U2 - 10.1007/s11432-019-9911-5

DO - 10.1007/s11432-019-9911-5

M3 - Article

AN - SCOPUS:85083079843

VL - 63

JO - Science in China, Series F: Information Sciences

JF - Science in China, Series F: Information Sciences

SN - 1009-2757

IS - 5

M1 - 152201

ER -