学术报告

题      目： Neural network models for the quaternion singular value decompositions

报  告  人：黄宝华   副教授  (邀请人：黎稳 )

                                福建师范大学

时      间：2022-11-03 15:00-16:00

腾 讯 会 议：816 654 599

报告人简介:

       黄宝华，福建师范大学副教授，主要研究方向为数值代数与数值优化，主持中国博士后科学基金面上项目一等资助，国家自然科学基金青年项目和福建省自然科学基金面上项目，在Numerical Linear Algebra with Applications、Journal of Global Optimization、Numerical Algorithms、Linear and Multilinaer Algebra等上发表论文 20 余篇。

摘      要：

       In this talk, we develop some neural network models for the computation of the quaternion singular value decomposition (QSVD). The considered QSVD neutral algorithms are shown to be the gradient flows on the product of Stiefel manifold over the quaternion field and their geometric and dynamical properties are established by using the differential geometry. Some numerical results confirm the feasibility and effectiveness of the proposed algorithms.