DOCSLIB.ORG

Stability Analysis of Positive Linear Systems by Decomposition of the State Matrices Into Symmetrical and Antisymmetrical Parts

Total Page：16

File Type：pdf, Size：1020Kb

Download full-text PDF Read full-text
Stability Analysis of Positive Linear Systems by Decomposition of the State Matrices Into Symmetrical and Antisymmetrical Parts BULLETIN OF THE POLISH ACADEMY OF SCIENCES TECHNICAL SCIENCES, Vol. 67, No. 4, 2019 DOI: 10.24425/bpasts.2019.130185 Stability analysis of positive linear systems by decomposition of the state matrices into symmetrical and antisymmetrical parts T. KACZOREK* Bialystok University of Technology, Faculty of Electrical Engineering, Wiejska 45D, 15-351 Bialystok Abstract. The stability of positive linear continuous-time and discrete-time systems is analyzed by the use of the decomposition of the state matrices into symmetrical and antisymmetrical parts. It is shown that: 1) The state Metzler matrix of positive continuous-time linear system is Hurwitz if and only if its symmetrical part is Hurwitz; 2) The state matrix of positive linear discrete-time system is Schur if and only if its symmetrical part is Hurwitz. These results are extended to inverse matrices of the state matrices of the positive linear systems. Key words: linear, positive, system, decomposition, state matrix, stability. 1. Introduction The following notation will be used: ℜ – the set of real n m n m numbers, ℜ £ – the set of n m real matrices, ℜ+£ – the set £ n n 1 A dynamical system is called positive if its trajectory starting of n m real matrices with nonnegative entries and ℜ = ℜ £ , £ + + from any nonnegative initial state remains forever in the posi- M – the set of n n Metzler matrices (real matrices with non- n £ tive orthant for all nonnegative inputs. An overview of state of negative off-diagonal entries), I – the n n identity matrix. n £ the art in positive systems theory is given in the monographs [1, 3, 7]. A variety of models having positive behavior can be found in engineering, especially in electrical circuits [15], eco- 2. Preliminaries nomics, social sciences, biology and medicine, etc. [3, 7]. The positivity and stability of linear systems have been Consider the continuous-time linear system investigated in [2, 5‒6, 8, 9, 16, 17, 22, 23] and of nonlin- ear systems in [10, 11]. A comparison of the stability of dis- x ̇ = Ax, (1) crete-time and continuous-time linear systems has been given n n n in [4]. The stability of interval positive linear systems with state where x = x(t) ℜ is the state vector and A = a ℜ £ . 2 ij 2 matrices in integer and rational powers has been analyzed in £ ¤ [14]. The linear systems have been intensively investigated in Definition 1. [3, 7] The system (1) is called (internally) positive [11, 12‒15, 18‒20] and the descriptor fractional linear systems if x(t) ℜn, t 0 for all x(0) ℜn. 2 + ¸ 2 + with different fractional orders in [21]. In this paper the asymptotic stability of positive continu- Theorem 1. [3, 7] The system (1) is positive if and only if ous-time and discrete-time linear systems by the decomposition of the state matrices into symmetrical and antisymmetrical parts A Mn. (2) will be addressed. 2 The paper is organized as follows. In Section 2 some prelim- Definition 2. [3, 7] The positive system (1) is called asymptot- inaries concerning positive continuous-time and discrete-time ically stable (the matrix A is called Hurwitz) if linear systems are recalled. Decomposition of the state matrices of positive linear continuous-time and discrete-time systems in n lim x(t) = 0 for all x(0) ℜ . (3) → + symmetrical and antisymmetrical parts and the stability of the t 1 2 symmetrical parts are addressed in Section 3. The Comparison of the stability of positive continuous-time and discrete-time Theorem 2. [3, 7] The positive system (1) is asymptotically linear systems is presented in Section 4. Some concluding stable if and only if one of the following equivalent conditions remarks are given in Section 5. is satisfied: 1. All coefficients of the characteristic polynomial * n n 1 e-mail: [email protected] det In s A = s + an 1 s ¡ + … + a1s + a0 (4) ¡ ¡ Manuscript submitted 2019-03-19, revised 2019-05-01 and 2019-05-23, £ ¤ initially accepted for publication 2019-06-19, published in August 2019 are positive, i.e. a > 0 for i = 0, 1, …, n 1. i ¡ Bull. Pol. Ac.: Tech. 67(4) 2019 761 T. Kaczorek – n n 2. All principal minors M , i = 1, …, n of the matrix –A are Theorem 6. [15] If A ℜ £ is Schur matrix then i 2 + positive, i.e. –1 n n (I A) ℜ £ . (14) – – – a11 – a12 n ¡ 2 + M1 = – a11 > 0, M2 = > 0, …, j j – a21 – a22 (5) Theorem 7. Let sk, k = 1, …, n be the eigenvalues of the matrix – n n M = det – A > 0. A ℜ £ . Then: n j j 2 1) –sk, k = 1, …, n are the eigenvalues of the matrix –A, £ ¤ T –1 3. There exists strictly positive vector λ = λ1 λn , λk > 0, 2) sk , k = 1, …, n are the eigenvalues of the inverse matrix ¢¢¢ –1 n n k = 1, …, n such that (det A = 0) A ℜ £ . £ ¤ 6 2 Aλ < 0 and λTA < 0. (6) Proof. From the equality Theorem 3. [15] If A M is asymptotically stable then –I I λ A = I (–λ) (– A) (15) 2 n n n ¡ n ¡ –1 n n £ ¤ £ ¤ –A ℜ £ . (7) it follows that det I (–λ) (– A) = 0 if and only if det I λ + n ¡ n ¡ 2 A = 0 since det – I (I λ A) = (–1)n det I λ A . ¡ £ n n ¡ ¤ n ¡ £ Consider the discrete-time linear system If det A = 0 then from the equality ¤ 6 £ ¤ £ ¤ x = Ax , i = 0, 1, … (8) –(Aλ)–1 I λ A = I λ–1 A–1 (16) i+1 1 n ¡ n ¡ n n –1 –1 where x is the state vector and A = a ℜ £ . it follows that det £I λ A¤ =£ 0 if and only¤ if det I λ i ij 2 n ¡ n ¡ A = 0 since £ ¤ ¡ £ ¤ £ Definition 3. [3, 7] The system (8) is called (internally) positive (Aλ)–1( λ A) = (Aλ)–1 λ A n n ¤det – In det – det In if xi ℜ+, i 0 for all x0 ℜ+. ¡ ¡ 2 ¸ 2 –1 and det Aλ = 0. □ £ 6 ¤ £ ¤ £ ¤ Theorem 4. The system (8) is positive if and only if £ ¤ Example 1. The matrix n n A ℜ £ . (9) 2 + –2 1 Definition 4. [3, 7] The positive system (8) is called asymptot- A = (17) ically stable (the matrix A is called Schur) if 1 –2 n lim x = 0 for all x ℜ . (10) with characteristic polynomial → i 0 + t 1 2 λ + 2 –1 2 Theorem 5. [3, 7] The positive system (8) is asymptotically det I2λ A = = λ + 4λ + 3 (18) stable if and only if one of the following equivalent conditions ¡ –1 λ + 2 £ ¤ is satisfied: j j 1) All coefficients of the characteristic polynomial has the eigenvalues λ1 = –1, λ2 = –3. The matrix n n 1 det In(z + 1) A = z + an 1 z ¡ + … + ¡ ¡ (11) 2 –1 det£In(z + 1) A¤ + a1z + a0 – A = (19) ¡ –1 2 £ ¤ are positive, i.e. a > 0 for i = 0, 1, …, n 1. i ¡ 2) All principal minors M ̂ , i = 1, …, n of the matrix I A with characteristic polynomial i n ¡ are positive, i.e. λ 2 –1 1 a – a det I λ (– A) = ¡ = λ2 + 4λ + 3 (20) ̂ ̂ ¡ 11 12 2 M1 = 1 a11 > 0, M2 = > 0, ¡ –1 λ 2 j ¡ j – a21 1 a22 ¡ (12) £ ¤ ¡ ̂ j j …, Mn = det In A > 0. j j ¡ has the eigenvalues λ1 = 1, λ2 = 3. £ ¤ The inverse matrix of (17) has the form 3) There exists strictly positive vector λ = λ λ T, λ > 0, 1 ¢¢¢ n k k = 1, …, n such that £ ¤ 1 –2 –1 A–1 = (21) Aλ < λ and λTA < λT. (13) 3 –1 –2 762 Bull. Pol. Ac.: Tech. 67(4) 2019 Stability analysis of positive linear systems by decomposition of the state matrices into symmetrical and antisymmetrical parts and its characteristic polynomial –9 + 33 –9 33 Note that the eigenvalues s = , s = ¡ of 1 2 2 2 – – 2 1 the matrix A and of the matrix As, s1, s2 are different and satisfy λ + the condition trA = trA = –9 = s– + s– . 3 3 4 1 s 1 2 det I λ A–1 = = λ2 + λ + (22) 2 ¡ 1 2 3 3 λ + Theorem 8. The Metzler matrix A Mn is Hurwitz if and only 3 3 2 £ ¤ if its symmetrical part As is Hurwitz. 1 has the zeros λ1 = –1, λ2 = – /3. Proof. By Theorem 2 the matrix A Mn is Hurwitz if and only j j 2 n This confirms the Theorem 7. if there exists a strictly positive vector λ ℜ such that 2 + Aλ < 0 and λTA < 0. (26) 3. Decomposition of the state matrices into T symmetrical and antisymmetrical parts Taking into account that (λTA) = ATλ < 0 and (1, 2) we obtain Consider the Metzler matrix A = a M which is in general ij 2 n A + AT Aλ + ATλ case not symmetrical. It is well-known the this matrix can be A λ = λ = < 0. (27) £ ¤ s 2 2 decomposed into symmetrical part As and the antisymmetrical part Aa Therefore, the Metzler matrix A is Hurwitz if and only if the A = As + Aa (23a) matrix As is Hurwitz. □ where Remark 1. If the matrix A = a M satisfies the conditions ij 2 n T T A + A A A n £ ¤ As = , Aa = ¡ (23b) 2 2 and ∑ aij < 0 for i = 1, …, n j =1 and T denotes the transpose. n (28) The following properties of the matrices (23b) are well- and aij < 0 for j = 1, …, n known: ∑ – – i =1 1) All eigenvalues s1, …, sn of the matrix As are real and of the matrix Aa = a ̂ ij , s ̂ 1, …, s ̂ n are imaginary.
Load more

Recommended publications
  • Stabilization, Estimation and Control of Linear Dynamical Systems with Positivity and Symmetry Constraints
    Stabilization, Estimation and Control of Linear Dynamical Systems with Positivity and Symmetry Constraints A Dissertation Presented by Amirreza Oghbaee to The Department of Electrical and Computer Engineering in partial fulfillment of the requirements for the degree of Doctor of Philosophy in Electrical Engineering Northeastern University Boston, Massachusetts April 2018 To my parents for their endless love and support i Contents List of Figures vi Acknowledgments vii Abstract of the Dissertation viii 1 Introduction 1 2 Matrices with Special Structures 4 2.1 Nonnegative (Positive) and Metzler Matrices . 4 2.1.1 Nonnegative Matrices and Eigenvalue Characterization . 6 2.1.2 Metzler Matrices . 8 2.1.3 Z-Matrices . 10 2.1.4 M-Matrices . 10 2.1.5 Totally Nonnegative (Positive) Matrices and Strictly Metzler Matrices . 12 2.2 Symmetric Matrices . 14 2.2.1 Properties of Symmetric Matrices . 14 2.2.2 Symmetrizer and Symmetrization . 15 2.2.3 Quadratic Form and Eigenvalues Characterization of Symmetric Matrices . 19 2.3 Nonnegative and Metzler Symmetric Matrices . 22 3 Positive and Symmetric Systems 27 3.1 Positive Systems . 27 3.1.1 Externally Positive Systems . 27 3.1.2 Internally Positive Systems . 29 3.1.3 Asymptotic Stability . 33 3.1.4 Bounded-Input Bounded-Output (BIBO) Stability . 34 3.1.5 Asymptotic Stability using Lyapunov Equation . 37 3.1.6 Robust Stability of Perturbed Systems . 38 3.1.7 Stability Radius . 40 3.2 Symmetric Systems . 43 3.3 Positive Symmetric Systems . 47 ii 4 Positive Stabilization of Dynamic Systems 50 4.1 Metzlerian Stabilization . 50 4.2 Maximizing the stability radius by state feedback .
    [Show full text]
  • On Control of Discrete-Time LTI Positive Systems
    Applied Mathematical Sciences, Vol. 11, 2017, no. 50, 2459 - 2476 HIKARI Ltd, www.m-hikari.com https://doi.org/10.12988/ams.2017.78246 On Control of Discrete-time LTI Positive Systems DuˇsanKrokavec and Anna Filasov´a Department of Cybernetics and Artificial Intelligence Faculty of Electrical Engineering and Informatics Technical University of Koˇsice Letn´a9/B, 042 00 Koˇsice,Slovakia Copyright c 2017 DuˇsanKrokavec and Anna Filasov´a.This article is distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. Abstract Incorporating an associated structure of constraints in the form of linear matrix inequalities, combined with the Lyapunov inequality guaranteing asymptotic stability of discrete-time positive linear system structures, new conditions are presented with which the state-feedback controllers can be designed. Associated solutions of the proposed design conditions are illustrated by numerical illustrative examples. Keywords: state feedback stabilization, linear discrete-time positive sys- tems, Schur matrices, linear matrix inequalities, asymptotic stability 1 Introduction Positive systems are often found in the modeling and control of engineering and industrial processes, whose state variables represent quantities that do not have meaning unless they are nonnegative [26]. The mathematical theory of Metzler matrices has a close relationship to the theory of positive linear time- invariant (LTI) dynamical systems, since in the state-space description form the system dynamics matrix of a positive systems is Metzler and the system input and output matrices are nonnegative matrices. Other references can find, e.g., in [8], [16], [19], [28]. The problem of Metzlerian system stabilization has been previously stud- ied, especially for single input and single output (SISO) continuous-time linear 2460 D.
    [Show full text]
  • Z Matrices, Linear Transformations, and Tensors
    Z matrices, linear transformations, and tensors M. Seetharama Gowda Department of Mathematics and Statistics University of Maryland, Baltimore County Baltimore, Maryland, USA [email protected] *************** International Conference on Tensors, Matrices, and their Applications Tianjin, China May 21-24, 2016 Z matrices, linear transformations, and tensors – p. 1/35 This is an expository talk on Z matrices, transformations on proper cones, and tensors. The objective is to show that these have very similar properties. Z matrices, linear transformations, and tensors – p. 2/35 Outline • The Z-property • M and strong (nonsingular) M-properties • The P -property • Complementarity problems • Zero-sum games • Dynamical systems Z matrices, linear transformations, and tensors – p. 3/35 Some notation • Rn : The Euclidean n-space of column vectors. n n • R+: Nonnegative orthant, x ∈ R+ ⇔ x ≥ 0. n n n • R++ : The interior of R+, x ∈++⇔ x > 0. • hx,yi: Usual inner product between x and y. • Rn×n: The space of all n × n real matrices. • σ(A): The set of all eigenvalues of A ∈ Rn×n. Z matrices, linear transformations, and tensors – p. 4/35 The Z-property A =[aij] is an n × n real matrix • A is a Z-matrix if aij ≤ 0 for all i =6 j. (In economics literature, −A is a Metzler matrix.) • We can write A = rI − B, where r ∈ R and B ≥ 0. Let ρ(B) denote the spectral radius of B. • A is an M-matrix if r ≥ ρ(B), • nonsingular (strong) M-matrix if r > ρ(B). Z matrices, linear transformations, and tensors – p. 5/35 The P -property • A is a P -matrix if all its principal minors are positive.
    [Show full text]
  • An Alternating Sequence Iteration's Method for Computing Largest Real Part Eigenvalue of Essentially Positive Matrices
    Computa & tio d n ie a l l Oepomo, J Appl Computat Math 2016, 5:6 p M p a Journal of A t h f DOI: 10.4172/2168-9679.1000334 e o m l a a n t r ISSN: 2168-9679i c u s o J Applied & Computational Mathematics Research Article Open Access An Alternating Sequence Iteration’s Method for Computing Largest Real Part Eigenvalue of Essentially Positive Matrices: Collatz and Perron- Frobernius’ Approach Tedja Santanoe Oepomo* Mathematics Division, Los Angeles Harbor College/West LA College, and School of International Business, California International University, USA Abstract This paper describes a new numerical method for the numerical solution of eigenvalues with the largest real part of essentially positive matrices. Finally, a numerical discussion is given to derive the required number of mathematical operations of the new method. Comparisons between the new method and several well know ones, such as Power and QR methods, were discussed. The process consists of computing lower and upper bounds which are monotonically approximating the eigenvalue. Keywords: Collatz’s theorem; Perron-Frobernius’ theorem; Background Eigenvalue Let A be an n x n essentially positive matrix. The new method can AMS subject classifications: 15A48; 15A18; 15A99; 65F10 and 65F15 be used to numerically determine the eigenvalue λA with the largest real part and the corresponding positive eigenvector x[A] for essentially Introduction positive matrices. This numerical method is based on previous manuscript [16]. A matrix A=(a ) is an essentially positive matrix if a A variety of numerical methods for finding eigenvalues of non- ij ij negative irreducible matrices have been reported over the last decades, ≥ 0 for all i ≠ j, 1 ≤ i, j ≤ n, and A is irreducible.
    [Show full text]
  • Matrix Algebra and Control
    Appendix A Matrix Algebra and Control Boldface lower case letters, e.g., a or b, denote vectors, boldface capital letters, e.g., A, M, denote matrices. A vector is a column matrix. Containing m elements (entries) it is referred to as an m-vector. The number of rows and columns of a matrix A is nand m, respectively. Then, A is an (n, m)-matrix or n x m-matrix (dimension n x m). The matrix A is called positive or non-negative if A>, 0 or A :2:, 0 , respectively, i.e., if the elements are real, positive and non-negative, respectively. A.1 Matrix Multiplication Two matrices A and B may only be multiplied, C = AB , if they are conformable. A has size n x m, B m x r, C n x r. Two matrices are conformable for multiplication if the number m of columns of the first matrix A equals the number m of rows of the second matrix B. Kronecker matrix products do not require conformable multiplicands. The elements or entries of the matrices are related as follows Cij = 2::;;'=1 AivBvj 'Vi = 1 ... n, j = 1 ... r . The jth column vector C,j of the matrix C as denoted in Eq.(A.I3) can be calculated from the columns Av and the entries BVj by the following relation; the jth row C j ' from rows Bv. and Ajv: column C j = L A,vBvj , row Cj ' = (CT),j = LAjvBv, (A. 1) /1=1 11=1 A matrix product, e.g., AB = (c: c:) (~b ~b) = 0 , may be zero although neither multipli­ cand A nor multiplicator B is zero.
    [Show full text]
  • LMI Based Principles in Strictly Metzlerian Systems Control Design
    Hindawi Mathematical Problems in Engineering Volume 2018, Article ID 9590253, 14 pages https://doi.org/10.1155/2018/9590253 Research Article LMI Based Principles in Strictly Metzlerian Systems Control Design Dušan Krokavec and Anna Filasová Department of Cybernetics and Artifcial Intelligence, Faculty of Electrical Engineering and Informatics, Technical University of Koˇsice, Letna9,04200Ko´ ˇsice, Slovakia Correspondence should be addressed to Duˇsan Krokavec; [email protected] Received 5 August 2017; Revised 15 January 2018; Accepted 22 April 2018; Published 17 July 2018 Academic Editor: Marzio Pennisi Copyright © 2018 Duˇsan Krokavec and Anna Filasova.´ Tis is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. Te paper is concerned with the design requirements that relax the existing conditions reported in the previous literature for continuous-time linear positive systems, reformulating the linear programming approach by the linear matrix inequalities principle. Incorporating an associated structure of linear matrix inequalities, combined with the Lyapunov inequality guaranteeing asymptotic stability of positive system structures, the conditions are presented, with which the state-feedback controllers and the system state observers can be designed. A numerical example illustrates the proposed conditions. 1. Introduction design. Te second group of methods represents [8], where LMI conditions to state observer design for positive linear Positive systems are ofen found in the modeling and control systems are presented, but the observer is not of Luenberger of engineering and industrial processes, whose state variables type since the standard observer gain matrix does not specify represent quantities that do not have meaning unless they the observer dynamics and an additive matrix is used to are nonnegative [1].
    [Show full text]
  • Optimising the Perron Eigenvalue with Applications to Dynamical Systems and Graphs
    Aleksandar Cvetkovic´ Gran Sasso Science Institute (GSSI) L'Aquila, Italy OPTIMISING THE PERRON EIGENVALUE WITH APPLICATIONS TO DYNAMICAL SYSTEMS AND GRAPHS PhD Thesis March 7, 2019 Thesis advisor: Vladimir Yu. Protasov University of L'Aquila, Italy Faculty of Computer Science of National Research University Higher School of Economics, Moscow, Russia G. Mahler - Symphony no. 10, Adagio Contents 1 Introduction 5 1.1 Thesis overview . .6 2 Positive product families 11 2.1 Simplex and greedy method: an overview . 12 2.2 Simplex vs. greedy method: a comparison . 14 3 Non-negative product families 17 3.1 Greedy method for non-negative families . 17 3.2 The (in)applicability of the greedy method . 21 3.3 Selective greedy method . 25 3.4 The convergence rate of the greedy method . 29 3.5 Selective greedy method: numerical results . 35 3.6 Optimizing spectral radius on reducible families . 39 4 Applications of the selective greedy method to non-negative matrices 41 4.1 Closest stable non-negative matrix . 42 4.2 Applications to graphs . 49 5 Metzler product families 59 5.1 Greedy method for full Metzler matrices . 61 5.2 Greedy method for sparse Metzler matrices . 64 5.3 Numerical results . 66 6 Closest (un)stable Metzler matrix 69 6.1 Hurwitz (de)stabilization in max-norm . 70 6.2 Hurwitz destabilization in L1 norm . 74 6.3 Hurwitz stabilization in L1 norm . 79 3 7 Applications of selective greedy method to Metzler matrices 85 7.1 Closest stable sign-matrix and its application to positive LSS . 87 Appendix: Implementation and strategies of the presented algo- rithms 97 Thesis summary 101 Bibliography 103 4 Chapter 1 Introduction Metzler matrices are an important tool for applied mathematics.
    [Show full text]
  • Arxiv:1907.07089V1 [Math.SP]
    UNIFYING MATRIX STABILITY CONCEPTS WITH A VIEW TO APPLICATIONS Olga Y. Kushel Shanghai University, Department of Mathematics, Shangda Road 99, 200444 Shanghai, China [email protected] Abstract Multiplicative and additive D-stability, diagonal stability, Schur D- stability, H-stability are classical concepts which arise in studying linear dynamical systems. We unify these types of stability, as well as many others, in one concept of (D, G, ◦)-stability, which depends on a stability region D ⊂ C, a matrix class G and a binary matrix operation ◦. This ap- proach allows us to unite several well-known matrix problems and to con- sider common methods of their analysis. In order to collect these methods, we make a historical review, concentrating on diagonal and D-stability. We prove some elementary properties of (D, G, ◦)-stable matrices, uniting the facts that are common for many partial cases. Basing on the proper- ties of a stability region D which may be chosen to be a concrete subset of C (e.g. the unit disk) or to belong to a specified type of regions (e.g. LMI regions) we briefly describe the methods of further development of the theory of (D, G, ◦)-stability. We mention some applications of the theory of (D, G, ◦)-stability to the dynamical systems of different types. Hurwitz stability, D-stability, diagonal stability, eigenvalue clustering, LMI regions, Lyapunov equation. MSC[2010] 15A48, 15A18, 15A75 Contents I Motivation and history 3 1 Introduction 3 1.1 OverviewofPaper .......................... 4 1.2 Unifyingconcept ........................... 5 1.3 Multiplicative (D, G)-stability.................... 6 arXiv:1907.07089v1 [math.SP] 14 Jul 2019 1.4 Hadamard (D, G)-stability.....................
    [Show full text]
  • New Initialization Strategy for Nonnegative Matrix Factorization
    NEW INITIALIZATION STRATEGY FOR NONNEGATIVE MATRIX FACTORIZATION A Thesis Presented By Yueyang Wang to The Department of Electrical and Computer Engineering In partial fulfillment of requirements for the degree of Master of Science in the field of Electrical & Computer Engineering Northeastern University Boston, Massachusetts May 2018 ii ABSTRACT Nonnegative matrix factorization (NMF) has been proved to be a powerful data representation method, and has shown success in applications such as data representation and document clustering. In this thesis, we propose a new initialization strategy for NMF. This new method is entitled square nonnegative matrix factorization, SQR-NMF. In this method, we first transform the non-square nonnegative matrix to a square one. Several strategies are proposed to achieve SQR step. Then we take the positive section of eigenvalues and eigenvectors for initialization. Simulation results show that SQR-NMF has faster convergence rate and provides an approximation with lower error rate as compared to SVD-NMF and random initialization methods. Complementing different elements in data matrix also affect the results. The experiments show that complementary elements should be 0 for small data sets and mean values of each row or column of the original nonnegative matrix for large data sets. Key words: Nonnegative matrix factorization, complementary elements iii ACKNOWLEDGEMENT I would like to express my sincere gratitude to my thesis advisor, Professor Shafai for his patience, motivation, and immense knowledge. His guidance helped me a lot in all the time of research and writing of the thesis. I could not have imagined having a better advisor for my master study. Besides my advisor, I would like to thank the rest of my thesis committee: Professor Nian Sun and Professor Vinay Ingle, for their insightful comments and encouragement.
    [Show full text]
  • LQ Problem in Stabilization of Linear Metzlerian Continuous-Time Systems
    INTERNATIONAL JOURNAL OF MATHEMATICAL MODELS AND METHODS IN APPLIED SCIENCES Volume 12, 2018 LQ Problem in Stabilization of Linear Metzlerian Continuous-time Systems D. Krokavec and A. Filasova represented by state-space equations, the synthesis of Abstract—The paper present a consistent set of linear matrix stabilizing state-feedback controllers, guaranteeing the closed- inequalities which guaranties asymptotic stability of the closed-loop loop system is asymptotically stable and internally positive, is system, warranties strictly Metzlerian system structure, and adjusts conditionally supported by linear programming to meet the the state and output variables coincident with prescribed quadratic closed-loop system positive structure [11], [12]. In order to limits. To realize with a positive control law gain, the diagonal stabilizability of strictly Metzlerian linear continuous-time systems is reduce the number of constraints entering the solution in linear approved, and the related closed-form expression of design condi- programming methods, an alternative synthesis procedure with tions is provided. The results are illustrated using a particular LQ is proposed in [13], where the system parameter boundaries problem, for which numerical examples are given. are defined by n linear matrix inequalities (LMI), if the system is strictly Metzlerian. Because a solution of such defined base Keywords—asymptotic stability, linear matrix inequalities, linear set of LMIs only assures that the closed-loop system matrix is quadratic control, Metzlerian continuous-time systems, state feedback strictly Metzler, the design conditions are complemented by stabilization. another LMI that imposes a stable asymptotic solution. Since the applied LMI variables are of diagonal matrix structure, it I. INTRODUCTION can be refereed about diagonal stabilizability of the strictly Positive systems indicate the processes whose variables Metzlerian continuous-time linear systems.
    [Show full text]
  • Perron–Frobenius Theorem - Wikipedia
    Perron–Frobenius theorem - Wikipedia https://en.wikipedia.org/wiki/Perron–Frobenius_theorem Perron–Frobenius theorem In linear algebra, the Perron–Frobenius theorem, proved by Oskar Perron (1907) and Georg Frobenius (1912), asserts that a real square matrix with positive entries has a unique largest real eigenvalue and that the corresponding eigenvector can be chosen to have strictly positive components, and also asserts a similar statement for certain classes of nonnegative matrices. This theorem has important applications to probability theory (ergodicity of Markov chains); to the theory of dynamical systems (subshifts of finite type); to economics (Okishio's theorem,[1] Hawkins–Simon condition[2]); to demography (Leslie population age distribution model);[3] to social networks (DeGroot learning process), to Internet search engines[4] and even to ranking of football teams.[5] The first to discuss the ordering of players within tournaments using Perron– Frobenius eigenvectors is Edmund Landau.[6] [7] Contents Statement Positive matrices Non-negative matrices Classification of matrices Perron–Frobenius theorem for irreducible matrices Further properties Applications Non-negative matrices Stochastic matrices Algebraic graph theory Finite Markov chains Compact operators Proof methods Perron root is strictly maximal eigenvalue for positive (and primitive) matrices Proof for positive matrices Lemma Power method and the positive eigenpair Multiplicity one No other non-negative eigenvectors Collatz–Wielandt formula Perron projection as a limit: Ak/rk Inequalities for Perron–Frobenius eigenvalue Further proofs 1 of 18 2/8/19, 14:03 Perron–Frobenius theorem - Wikipedia https://en.wikipedia.org/wiki/Perron–Frobenius_theorem Perron projection Peripheral projection Cyclicity Caveats Terminology See also Notes References Original papers Further reading Statement Let positive and non-negative respectively describe matrices with exclusively positive real numbers as elements and matrices with exclusively non-negative real numbers as elements.
    [Show full text]
  • Stability Analysis of Positive Systems Vahid Samadi Bokharaie
    Stability Analysis of Positive Systems with Applications to Epidemiology A dissertation submitted for the degree of Doctor of Philosophy By: Vahid Samadi Bokharaie Under the supervision of: Dr. Oliver Mason Hamilton Institute National University of Ireland Maynooth Ollscoil na hÉireann, Má Nuad January 2012 To all those who offered me their help and support, in my ongoing journey, to become who I want to be. i Contents Contents ii 1 Introduction1 1.1 Introductory Remarks ...................... 1 1.2 Overview.............................. 5 1.3 Publications............................ 7 2 Background8 2.1 Introduction............................ 8 2.2 Basic Concepts in Matrix Analysis ............... 9 2.3 KKM Lemma........................... 10 2.4 Dynamical Systems........................ 12 2.4.1 Stability.......................... 14 2.4.2 Cooperative Systems................... 15 2.4.3 Positive Systems ..................... 17 2.5 Positive LTI Systems....................... 18 2.6 Positive Nonlinear Systems.................... 24 2.6.1 Homogeneous Cooperative Systems........... 26 2.6.2 Subhomogeneous Cooperative Systems......... 29 3 Stability Conditions for Positive Nonlinear Systems 33 3.1 Introduction............................ 33 3.2 Background............................ 34 3.3 Literature Review......................... 37 3.4 D-stability for General Cooperative Systems.......... 40 3.4.1 Unique Equilibrium at the Origin............ 40 3.4.2 Unique Equilibrium in the Interior of the Positive Orthant 44 3.5 D-stability for Homogeneous Cooperative Systems . 48 ii Contents 3.6 D-stability for Subhomogeneous Cooperative Systems . 53 3.6.1 Equilibrium at the origin................. 53 3.6.2 Equilibrium in the interior of the Positive Orthant . 56 3.7 D-stability for Planar Cooperative Systems........... 57 3.8 Positivity of Equilibria...................... 61 3.9 Concluding Remarks......................
    [Show full text]