Application of gradient descent algorithms based on geodesic distances

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.

[1]  Jang-Won Lee,et al.  Stochastic utility-based flow control algorithm for services with time-varying rate requirements , 2012, Comput. Networks.

[2]  A. Guven.  Approximation of Continuous Functions by Matrix Means of Hexagonal Fourier Series , 2018 .

[3]  N. Yoshida,et al.  SIMULATIONS OF BARYON ACOUSTIC OSCILLATIONS. II. COVARIANCE MATRIX OF THE MATTER POWER SPECTRUM , 2009, 0902.0371.

[4]  R. Adamczak,et al.  Quantitative estimates of the convergence of the empirical covariance matrix in log-concave ensembles , 2009, 0903.2323.

[5]  Toshihisa Tanaka,et al.  Empirical Arithmetic Averaging Over the Compact Stiefel Manifold , 2013, IEEE Transactions on Signal Processing.

[6]  Toshihisa Tanaka,et al.  An Algorithm to Compute Averages on Matrix Lie Groups , 2009, IEEE Transactions on Signal Processing.

[7]  Marc Arnaudon,et al.  Barycenters of measures transported by stochastic flows , 2005 .

[8]  Bruno Iannazzo,et al.  The Riemannian Barzilai–Borwein method with nonmonotone line search and the matrix geometric mean computation , 2018 .

[9]  Tsuyoshi Murata,et al.  {m , 1934, ACML.

[10]  Jianping Li,et al.  Normalized natural gradient in independent component analysis , 2010, Signal Process..

[11]  Simone G. O. Fiori,et al.  Learning the Fréchet Mean over the Manifold of Symmetric Positive-Definite Matrices , 2009, Cognitive Computation.

[12]  QU Long-hai,et al.  Application of Information Geometry to Target Detection for Pulsed-doppler Radar , 2011 .

[13]  Maher Moakher.  On the Averaging of Symmetric Positive-Definite Tensors , 2006 .

[14]  Xian-Da Zhang,et al.  Matrix Analysis and Applications , 2017 .

[15]  R. Bhatia,et al.  Inequalities for the Wasserstein mean of positive definite matrices , 2018, Linear Algebra and its Applications.

[16]  Maher Moakher,et al.  A Differential Geometric Approach to the Geometric Mean of Symmetric Positive-Definite Matrices , 2005, SIAM J. Matrix Anal. Appl..

[17]  H. Karcher,et al.  Jacobi fields and Finsler metrics on compact Lie groups with an application to differentiable pinching problems , 1974 .

[18]  Chunhui Li,et al.  Optimal control on special Euclidean group via natural gradient algorithm , 2015, Science China Information Sciences.

[19]  Lin Zhao,et al.  An Adaptive Unscented Kalman Filtering Algorithm for MEMS/GPS Integrated Navigation Systems , 2014, J. Appl. Math..

[20]  Xiaomin Duan,et al.  Riemannian Gradient Algorithm for the Numerical Solution of Linear Matrix Equations , 2014, J. Appl. Math..

[21]  Junsheng Zhao,et al.  Adaptive natural gradient learning algorithms for Mackey-Glass chaotic time prediction , 2015, Neurocomputing.

[22]  H. Karcher.  Riemannian center of mass and mollifier smoothing , 1977 .

[23]  Dario Bini,et al.  Computing the Karcher mean of symmetric positive definite matrices , 2013 .

[24]  Frédéric Barbaresco,et al.  Interactions between Symmetric Cone and Information Geometries: Bruhat-Tits and Siegel Spaces Models for High Resolution Autoregressive Doppler Imagery , 2009, ETVC.

[25]  C S Hughes,et al.  Anticipating the next meal using meal behavioral profiles: A hybrid model-based stochastic predictive control algorithm for T1DM , 2011, Comput. Methods Programs Biomed..

[26]  Linyu Peng,et al.  A Natural Gradient Algorithm for Stochastic Distribution Systems , 2014, Entropy.

[27]  Shun-ichi Amari,et al.  Why natural gradient? , 1998, Proceedings of the 1998 IEEE International Conference on Acoustics, Speech and Signal Processing, ICASSP '98 (Cat. No.98CH36181).

[28]  Rachid Deriche,et al.  Statistics on the Manifold of Multivariate Normal Distributions: Theory and Application to Diffusion Tensor MRI Processing , 2006, Journal of Mathematical Imaging and Vision.

[29]  Maher Moakher,et al.  Means of Hermitian positive-definite matrices based on the log-determinant α-divergence function , 2012 .

[30]  F. Barbaresco.  Innovative tools for radar signal processing Based on Cartan’s geometry of SPD matrices & Information Geometry , 2008, 2008 IEEE Radar Conference.

[31]  Elham Nobari,et al.  A geometric mean for Toeplitz and Toeplitz-block block-Toeplitz matrices , 2018, Linear Algebra and its Applications.

[32]  Haluk Gozde,et al.  Comparative performance analysis of Artificial Bee Colony algorithm in automatic generation control for interconnected reheat thermal power system , 2012 .

[33]  Shun-ichi Amari,et al.  Natural Gradient Works Efficiently in Learning , 1998, Neural Computation.

[34]  J. Jost.  Riemannian geometry and geometric analysis , 1995 .