The Exact Distribution of the Condition Number of Complex Random ...

4 downloads 0 Views 485KB Size Report
Sep 24, 2013 - The Exact Distribution of the Condition Number of. Complex Random Matrices. Lin Shi,1 Taibin Gan,2 Hong Zhu,1 and Xianming Gu2. 1 SchoolΒ ...
Hindawi Publishing Corporation The Scientific World Journal Volume 2013, Article ID 729839, 4 pages http://dx.doi.org/10.1155/2013/729839

Research Article The Exact Distribution of the Condition Number of Complex Random Matrices Lin Shi,1 Taibin Gan,2 Hong Zhu,1 and Xianming Gu2 1 2

School of Automation Engineering, University of Electronic Science and Technology of China, Chengdu, Sichuan 611731, China School of Mathematical Sciences, University of Electronic Science and Technology of China, Chengdu, Sichuan 611731, China

Correspondence should be addressed to Lin Shi; [email protected] Received 19 August 2013; Accepted 24 September 2013 Academic Editors: E. A. Abdel-Salam and E. Francomano Copyright Β© 2013 Lin Shi et al. This 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. 𝐻 πΊπ‘šΓ—π‘› which is the complex Wishart matrix. Let πœ† 1 > πœ† 2 > . . . > πœ† 𝑛 > 0 Let πΊπ‘šΓ—π‘› (π‘š β‰₯ 𝑛) be a complex random matrix and π‘Š = πΊπ‘šΓ—π‘› and 𝜎1 > 𝜎2 > . . . > πœŽπ‘› > 0 denote the eigenvalues of the W and singular values of πΊπ‘šΓ—π‘› , respectively. The 2-norm condition number of πΊπ‘šΓ—π‘› is πœ…2 (πΊπ‘šΓ—π‘› ) = βˆšπœ† 1 /πœ† 𝑛 = 𝜎1 /πœŽπ‘› . In this paper, the exact distribution of the condition number of the complex Wishart matrices is derived. The distribution is expressed in terms of complex zonal polynomials.

1. Introduction Over the past decade, multiple-input and multiple-output (MIMO) systems have been at the forefront of wireless communications research and development, due to their huge potential for delivering significant capacity compared with conventional systems [1–13]. The capacity and performance of practical MIMO transmission schemes are often dictated by the statistical eigenproperties of the instantaneous channel 𝐻 πΊπ‘šΓ—π‘› , where πΊπ‘šΓ—π‘› is a complex correlation matrix π‘Š = πΊπ‘šΓ—π‘› Gaussian matrix and π‘Š is known to follow a complex Wishart distribution. In recent years, the statistical properties of Wishart matrices have been extensively studied and applied to a large number of MIMO applications. In statistics, the random eigenvalues are used in hypothesis testing, principal component analysis, canonical correlation analysis, multiple discriminant analysis, and so forth (see [7]). In nuclear physics, random eigenvalues are used to model nuclear energy levels and level spacing [6]. Moreover, the zeros of the Riemann zeta function are modeled using random eigenvalues [6]. Condition numbers arise in theory and applications of random matrices, such as multivariate statistics and quantum physics. Let πΊπ‘šΓ—π‘› (π‘š β‰₯ 𝑛) be a complex random matrix whose elements are independent and identically distributed (i.i.d.)

standard normal random variables. As we know, the 𝑛 Γ— 𝐻 πΊπ‘šΓ—π‘› is a complex 𝑛 complex random matrix π‘Š = πΊπ‘šΓ—π‘› Wishart matrix. Its distribution is denoted by π‘ŠβˆΌπ‘Š(π‘š, Ξ£), Ξ£ = 𝜎2 𝐼. In addition, π‘Š is a positive definite Hermitian matrix with real eigenvalues; let πœ† 1 > πœ† 2 > β‹… β‹… β‹… > πœ† 𝑛 > 0 and 𝜎1 > 𝜎2 > β‹… β‹… β‹… > πœŽπ‘› > 0 denote the eigenvalues of the π‘Š and singular values of πΊπ‘šΓ—π‘› , respectively. The 2-norm condition number of πΊπ‘šΓ—π‘› is πœ…2 (πΊπ‘šΓ—π‘› ) = βˆšπœ† 1 /πœ† 𝑛 = 𝜎1 /πœŽπ‘› ; thus, πœ…2 (π‘Š) = πœ…2 (πΊπ‘šΓ—π‘› )2 . The exact distributions of the condition number of a 2 Γ— 𝑛 matrix whose elements are independent and identically distributed standard normal real or complex random variables are given in [5] by Edelman. Edelman also obtained the limiting distributions and the limiting expected logarithms of the condition numbers of random rectangular matrices whose elements are independent and identically distributed standard normal random variables. The exact distributions of the condition number of Gaussian matrices are studied in [1] for real random matrix. Here, we derive the exact distribution of the condition number of a complex random matrix for special case Ξ£ = 𝜎2 𝐼. This paper is arranged as follows. Section 2 gives some preliminary results to the complex random matrices. In Section 3, the main result of this work, the density function of πœ…2 (π‘Šπ‘›Γ—π‘› ) for complex case, is proved.

2

The Scientific World Journal

2. Some Preliminary Results

where

In this section, we give some results on joint density of the eigenvalues of a complex Wishart matrix π‘Š and π‘ŠβˆΌπ‘Š(π‘š, Ξ£), Ξ£ = 𝜎2 𝐼. The determinant, trace, and norm of a square matrix 𝐡 are denoted by |𝐡|, tr(𝐡), and ‖𝐡‖, respectively. For any nonnegative integer 𝑝, a portion of 𝑝 is a multiple (π‘Ž1 , π‘Ž2 , . . . , π‘Žπ‘› ), where π‘Ž1 β‰₯ π‘Ž2 β‰₯ β‹… β‹… β‹… β‰₯ π‘Žπ‘› β‰₯ 0 such that βˆ‘π‘›π‘–=1 π‘Žπ‘– = 𝑝, D𝑝 is the set of all portions of 𝑝, and the symbol βˆ‘π‘ π‘Žπœ… means the summation over D𝑝 ; that is, βˆ‘π‘ π‘Žπœ… = βˆ‘πœ…βˆˆD𝑝 π‘Žπœ… . Let πœ… be any portion of 𝑝, and let πœ‡1 > πœ‡2 > β‹… β‹… β‹… > πœ‡π‘› be the eigenvalues of an 𝑛 Γ— 𝑛 matrix 𝐡 as follows: 𝛼[πœ…] = 𝑝!

βˆπ‘›π‘– 0 be the eigenvalues of π‘Š and Ξ› = diag(πœ† 1 , πœ† 2 , . . . , πœ† 𝑛 ). If π‘ŠβˆΌπ‘Š(π‘š, Ξ£), Ξ£ = 𝜎2 𝐼 with π‘š β‰₯ 𝑛, then the joint density of its eigenvalues is defined [2] as follows:

𝑛 βˆπœ†π‘šβˆ’π‘› ∏(πœ† 𝑖 𝑖 𝑖=1 𝑖 πœ‰2 > β‹… β‹… β‹… > πœ‰π‘› > 0}, then [2]

𝑛

1 1 [ 𝑛] = ∏( (𝑛 βˆ’ 𝑖 + 1)) , 2 πœ… 𝑖=1 2 π‘Žπ‘–

π‘šβˆ’π‘›

𝜏 where 𝑑 = 𝑠 + 𝑝 and 𝑔],πœ… is a constant coefficient.

where

𝑛

(8)

𝜏∈D𝑑

𝜏 where 𝑑 = 𝑠 + 𝑝 and 𝑔],πœ… is a constant coefficient.

The following is the basic properties of the zonal polynomials [8]:

𝑓 (Ξ›) =

𝑑π‘₯βˆ’1 π‘’βˆ’π‘‘ 𝑑𝑑.

Lemma 2. For an 𝑛 Γ— 𝑛 matrix 𝐡, the product of two zonal polynomials can be expressed in terms of a weighted combination of another zonal polynomial [7]; that is, for all ] ∈ D𝑠 and βˆ€πœ… ∈ D𝑝 , one has

(1)

πΆπœ… (𝐡) = 𝛼[πœ…] β‹… 𝛼[πœ…] (𝐡)

πœ‹π‘›(π‘›βˆ’1) (𝜎2 )

(7)

𝜏 𝐢] (𝐡) πΆπœ… (𝐡) = βˆ‘ 𝑔],πœ… 𝐢𝜏 (𝐡) ,

The zonal polynomials (also called Schur polynomials) of 𝐡 are defined by

(π‘₯)π‘Žπ‘– = π‘₯ (π‘₯ + 1) β‹… β‹… β‹… (π‘₯ + π‘Žπ‘– βˆ’ 1) ,

+∞

0

,

󡄨 󡄨󡄨 󡄨󡄨[(πœ‡π‘Žπ‘— +π‘›βˆ’π‘— )]󡄨󡄨󡄨 󡄨 󡄨󡄨 𝑖 π›Όπœ… (𝐡) = 󡄨 π‘›βˆ’π‘— 󡄨 󡄨 . 󡄨󡄨[(πœ‡ )]󡄨󡄨 󡄨󡄨 𝑖 󡄨󡄨

𝑝

𝑛

Γ𝑛 (π‘Ž) = πœ‹π‘›(π‘›βˆ’1)/2 βˆΞ“ (π‘Ž + π‘˜ + 1) ,

1 𝑛 βˆ’ πœ† 𝑙 ) exp (βˆ’ 2 βˆ‘πœ† 𝑖 ) , 𝜎 𝑖=1 2

(6)

The exact distribution of the 2-norm condition number of the Wishart matrix π‘Š is derived by the following. Theorem 5. Let πΊπ‘šΓ—π‘› be a complex random matrix and π‘Š = 𝐻 πΊπ‘šΓ—π‘› . πœ† 1 and πœ† 𝑛 are the maximum and minimum πΊπ‘šΓ—π‘›

The Scientific World Journal

3

eigenvalues of π‘Š, πœ…2 (π‘Š) = πœ† 1 /πœ† 𝑛 . Then, the exact distribution of πœ…2 (π‘Š) is given by βˆ’π‘šπ‘›

πœ‹π‘›(π‘›βˆ’1) (𝜎2 )

𝑓 (πœ…2 (π‘Š)) =

Γ𝑛 (𝑛) Γ𝑛 (π‘š)

exp (βˆ’

𝑛 πœ† ) 𝜎2 1

𝑝=0 πœ…βˆˆD𝑝

1>π›Ύπ‘›βˆ’1 >β‹…β‹…β‹…>𝛾2 >0

∞ πœ†1 1 βˆ‘ βˆ‘ βˆ‘ (𝑛 βˆ’ π‘š)] 𝜎2𝑝 𝑝! 𝑠=0 ]∈D 𝜏∈D 𝑠! 𝑠

𝑛

∫

π‘šπ‘›βˆ’π‘›+𝑝

∞

Γ—βˆ‘ βˆ‘

By making the transformation πœ‰π‘– = 𝛾𝑖 /𝛾𝑛 , 𝑖 = 2, . . . , 𝑛 βˆ’ 1, using Lemma 3, and integrating over the set 1 > πœ‰π‘›βˆ’1 > β‹… β‹… β‹… > πœ‰2 > 0, we have

(π‘›βˆ’1)(𝑛+1)+𝑝+π‘ βˆ’1

Γ— πœ…2 (π‘Š)βˆ’(π‘›βˆ’1)(𝑛+1)βˆ’π‘βˆ’π‘ +1 Γ— (𝑛2 βˆ’ 1 + 𝑝 + 𝑠)

(𝜎 )

exp (βˆ’

Γ𝑛 (𝑛) Γ𝑛 (π‘š)

1>πœ‰π‘›βˆ’1 >β‹…β‹…β‹…>πœ‰2 >0

𝑛

𝑛 πœ† ) πœ†π‘šπ‘›βˆ’π‘› |𝐴|2 |𝐼 βˆ’ 𝐴|π‘šβˆ’π‘› 𝜎2 1 1

2

Γ— ∏(π›Ύπ‘˜ βˆ’ π›Ύπ‘˜ ) exp [ π‘˜πœ‰π‘›βˆ’1 >β‹…β‹…β‹…>πœ‰2 >0

πœ‹

(𝜎 )

Γ𝑛 (𝑛) Γ𝑛 (π‘š) 𝑛

2

Γ— ∏(𝛾𝑖 βˆ’ 𝛾𝑙 ) βˆ‘ βˆ‘ 𝑖 β‹… β‹… β‹… > 𝛾2 > 0, we obtain the joint distribution of πœ† 1 and 𝛾2 , . . . , 𝛾𝑛 , as follows: πœ‹

Γ—βˆ«

Γ— ∏(1 βˆ’ πœ‰π‘– ) ∏ (πœ‰π‘– βˆ’ πœ‰π‘— ) βˆπ‘‘πœ‰π‘– ,

where 𝑑 = 𝑠 + 𝑝.

2 βˆ’π‘šπ‘›

𝑖=2

𝑖