孫曉祥,杜宇靜
(吉林農(nóng)業(yè)科技學(xué)院 文理學(xué)院,吉林 吉林 132101)
?
研究簡(jiǎn)報(bào)
分塊相鄰隨機(jī)矩陣最大特征值的極限性質(zhì)
孫曉祥,杜宇靜
(吉林農(nóng)業(yè)科技學(xué)院 文理學(xué)院,吉林 吉林 132101)
利用隨機(jī)矩陣的矩方法和譜分析理論研究分塊相鄰隨機(jī)矩陣最大特征值的極限,在一定矩條件假設(shè)下,得到了該矩陣最大特征值上極限的界.
分塊矩陣;相鄰矩陣;隨機(jī)矩陣;最大特征值
分塊矩陣在物理、圖論和無(wú)線電通訊等領(lǐng)域應(yīng)用廣泛,目前已有許多研究成果[1-13].通常固定的分塊結(jié)構(gòu)是預(yù)先給定的,而子塊一般假設(shè)是Wigner矩陣、Toeplitz矩陣和Hankel矩陣.文獻(xiàn)[1-2]研究了當(dāng)子塊為Wigner矩陣時(shí)分塊矩陣的性質(zhì);文獻(xiàn)[3-6]研究了當(dāng)分塊結(jié)構(gòu)為Toepliz矩陣時(shí)分塊矩陣的相關(guān)性質(zhì).當(dāng)矩陣元素為高斯分布時(shí),Far等[12]利用運(yùn)算值的自由概率定理建立了這種分塊矩陣極限譜分布的性質(zhì).本文利用矩方法在矩陣元素滿足一定的矩條件假設(shè)下,考慮子塊為矩形矩陣情況下分塊相鄰隨機(jī)矩陣最大特征值的極限性質(zhì).
定義n×n矩陣如下:
定理1在假設(shè)(H1)~(H3)下,有
(1)
用矩方法證明定理1.首先,將矩陣元素截?cái)?有如下命題.
命題1如果不等式(1)在下列條件下成立,則其在定理1的假設(shè)下也成立:
對(duì)任意的1≤r
λ1(A+B)≤λ1(A)+λ1(B).
因此,
注意到
根據(jù)Markov不等式,對(duì)任意的ε>0,有
(2)
因此可以假設(shè)Mn的對(duì)角線元素全為零.選擇θn=ηn/2,令
由Markov不等式,有
(3)
從而有
(4)
綜上有
(5)
易驗(yàn)證:
此外,
因此
選擇一列偶數(shù)k=2[(logn)2],這里[(logn)2]表示(logn)2的整數(shù)部分.
(6)
由命題1的假設(shè)條件2)和3),可斷定G(r,t)的每項(xiàng)不超過(guò)σ2rbt(ηnn1/2)k-2r-t.于是
(7)
經(jīng)過(guò)計(jì)算可知
(8)
對(duì)式(8)最右端應(yīng)用如下不等式(取a=t+1):
可知式(7)最右端有上界:
由于ηn→0,故有
從而
(9)
又因?yàn)閗/logn→∞,式(9)是絕對(duì)可和的,所以利用Borel-Cantelli引理可知定理1的結(jié)論成立.
[1] Oraby T.The Spectral Laws of Hermitian Block-Matrices with Large Random Blocks [J].Electron Commun Probab,2007,12:465-476.
[2] Banerjee S,Bose A.Noncrossing Partitions,Catalan Words and the Semicircular Law [J].J Theoret Probab,2013,26(2):386-409.
[3] Gazzah H,Regalia P A,Delmas J P.Asymptotic Eigenvalue Distribution of Block Toeplitz Matrices and Application to Blind SIMO Channel Identification [J].IEEE Trans Inform Theory,2001,47(3):1243-1251.
[4] LI Yiting,LIU Dangzheng,WANG Zhengdong.Limit Distributions of Eigenvalues for Random Block Toeplitz and Hankel Matrices [J].J Theoret Probab,2011,24(4):1063-1086.
[5] Basu R,Bose A,Ganguly S,et al.Limiting Spectral Distribution of Block Matrices with Toeplitz Block Structure [J].Stat Probabil Lett,2012,82(7):1430-1438.
[6] Far R R,Oraby T,Bryc W,et al.On Slow-Fading MIMO Systems with Nonseparable Correlation [J].IEEE Trans Inform Theory,2008,54(2):544-553.
[7] Girko V L.Random Block Matrix Density and SS-Law [J].Random Oper Stoch Equ,2000,8(2):189-194.
[8] Müller R R.On the Asymptotic Eigenvalue Distribution of Concatenated Vector-Valued Fading Channels [J].IEEE Trans Inf Theory,2002,48(7):2086-2091.
[9] Bolla M.Distribution of the Eigenvalues of Random Block-Matrices [J].Lin Alg Appl,2004,377:219-240.
[10] Cottatellucci L,Müller R R.CDMA Systems with Correlated Spatial Diversity:A Generalized Resource Pooling Result [J].IEEE Trans Inf Theory,2007,53(3):1116-1136.
[11] Dette H,Reuther B.Random Block Matrices and Matrix Orthogonal Polynomials [J].J Theoret Probab,2010,23(2):378-400.
[12] Far R R,Oraby T,Bryc W,et al.Spectra of Large Block Matrices [J/OL].2006-10-09.http://xxx.lanl.gov/abs/CSIT/0610045.
[13] BAI Zhidong,Silverstein J W.Spectral Analysis of Large Dimensional Random Matrices [M].New York:Springer-Verlag,2010:100-101.
(責(zé)任編輯:趙立芹)
LimitPropertiesoftheBlockRandomAdjacencyMatrix
SUN Xiaoxiang,DU Yujing
(SchoolofArtandScience,JilinAgriculturalScienceandTechnologyCollege,Jilin132101,JilinProvince,China)
We studied the limit of the largest eigenvalue of the block adjacency matrix using the moment method and spectral theory in random matrix theory.Under some moments assumptions,we obtained a bound of the super limit of the largest eigenvalue.
block matrix;adjacency matrix;random matrix;largest eigenvalue
10.13413/j.cnki.jdxblxb.2015.03.21
2014-10-13.
孫曉祥(1967—),男,滿族,碩士,副教授,從事概率論與數(shù)理統(tǒng)計(jì)的研究,E-mail:jlnkusxx@163.com.通信作者:杜宇靜(1969—),女,漢族,博士,教授,從事概率論與數(shù)理統(tǒng)計(jì)的研究,E-mail:duyj219@163.com.
國(guó)家自然科學(xué)基金(批準(zhǔn)號(hào):11471068).
O211.4
:A
:1671-5489(2015)03-0461-04