December 6, 2025
Saeed Karimi

Saeed Karimi

Academic Rank: Associate professor
Address:
Degree: Ph.D in Applied Mathematics
Phone: 07733447965
Faculty: Faculty of Intelligent Systems and Data Science

Research

Title
Iterative methods for solving constrained tensor equations
Type Thesis
Keywords
تانسور، معادلات تانسوري مقيد، كمترين مربعات تعميم يافته، كمترين مربعات تعميم يافته كلي، ضرب انيشتين
Researchers bentohoda zali (Student) , Saeed Karimi (First primary advisor) ,

Abstract

This dissertation presents novel and efficient tensor­based methods for solving constrained tensor equations and generalized least squares problems. Initially, constrained tensor equations along with their concepts and applications are described. Inspired by iterative algorithms for solving constrained matrix equations, an iterative algorithm based on the generalized least squares method is proposed, utilizing Einstein summation for solving constrained tensor equations. This algorithm is capable of addressing tensor equations with general constraints, and by considering the tensor structure, it aids in reducing computational complexity. To implement the generalized least squares algorithm efficiently in the tensor space, multilinear operators and their duals are defined and analyzed. The stability and convergence analysis of the proposed method is thoroughly presented, demonstrating strong mathematical properties and reliability of the technique. Furthermore, a novel approach for solving the generalized least squares problem is introduced. This problem is an extension of the generalized least squares problem under conditions where overdetermined linear systems AX ≈ B face significant errors in both the data matrix A and the observation vector B. Leveraging first­order Taylor series expansion, the generalized least squares problem is transformed into a linear problem, allowing for the application of the generalized least squares algorithm in a tensor framework. This transformation leads to simplification of computations, improved accuracy of results, and reduced execution time. The efficiency and accuracy of the proposed methods are validated through performance evaluations in the context of image reconstruction as a practical application. Results obtained from numerous numerical examples, alongside comparisons with existing methods in reputable literature, demonstrate the superiority of the proposed methods in terms of computational efficiency and accuracy of res