An Upper Bound for Limited Rate Feedback MIMO Capacity
|
|
- Teresa Grant
- 5 years ago
- Views:
Transcription
1 1 A Upper Boud for Limited Rate Feedbac MIMO Capacity Göha M Güvese, Studet Member, IEEE, ad A Özgür Yılmaz Abstract We develop a techique to upper boud the poitto-poit MIMO limited rate feedbac LRF capacity uder a wide class of vector quatizatio schemes The upper boud turs out to be tight ad ca also be used to obtai a absolute upper boud by usig a boudig distributio for Grassmaia beamformig The boudig techique ca be applied to other problems requirig the exact evaluatio of the expected value of matrix determiat Idex Terms Limited rate feedbac, MIMO, capacity boud, boudig distributio, vector quatizatio, reduced precodig, sigular value decompositio I INTRODUCTION Capacity gais promised by multi-iput multi-output MIMO systems ofte require a accurate owledge of the chael at trasmitter ad receiver sides especially i quest to capitalize these possible gais i practical systems A accuracy problem arises whe chael state iformatio CSI has to be trasmitted from the receiver to the trasmitter It is obvious that CSI caot be trasmitted with ifiite precisio A limited rate feedbac chael is usually available for this commuicatio ad this sets a limit for the accuracy of CSI at the trasmitter side It was show that the MIMO chael is iterferece limited whe the chael estimatio is imperfect [1] It was further observed i [1] that istataeous feedbac, eve if imperfect, gives large capacity gais i low SNR ad is useful i high SNR especially whe the umber of trasmit ateas t is larger tha the that of receive ateas r [] I [3], quatizatio rules ad correspodig quatizer desig criteria were proposed to be used i MISO multiple-iput sigleoutput ad MIMO chaels Quatizatio of beamformers were ivestigated uder a Grassmaia lie pacig framewor with regard to quatizatio codeboo size, capacity-snr loss, ad outage performace i [4], [5] We ivestigate the capacity of poit-to-poit MIMO chaels i this paper as opposed to the broadcast chael settigs i aforemetioed studies [6] Although the capacity is less affected by the lac of CSI o the trasmitter side at high SNR [7], its availability is very importat both at low SNR ad i desigig practical systems that ca operate close to the capacity as i adaptively modulated MIMO schemes [8] sice Mauscript received December, 7; revised May 1, 8; accepted July 1, 8 The associate editor coordiatig the review of this paper ad approvig it for publicatio was Dr Rohit Nabar This wor was supported i part by the Scietific ad Techological Research Coucil of Turey TUBITAK uder grat 14E7 The authors are with the Departmet of Electrical ad Electroics Egieerig, Middle East Techical Uiversity, Aara, Turey aoyilmaz@metuedutr, guvese@metuedutr the complex tas of joit detectio ad decodig is avoided Furthermore, the capacity is strictly smaller with t > r if o CSI is available at trasmitter [] We cocetrate o a fiite rate feedbac sceario i which precoders obtaied by the sigular value decompositio of the MIMO chael [9] are fed bac to the trasmitter side A capacity loss boud for covariace matrix based quatizatio was preseted i [1] ad a capacity loss boud was proposed i [11] for desigig matrix quatizatio based codeboos We herei focus o quatizig the colums of the precodig matrix obtaied from sigular value decompositio SVD The chael is quasi-parallelized by separately quatizig precoders ad well-ow adaptive modulatio ad codig techiques ca be utilized as stated i [] Covariace matrices geerated radomly with uiform distributio o the uit sphere are used i [1], that is, radom matrix quatizatio is studied O the other had, our mai cotributio i this paper is the derivatio of a capacity upper boud expressio that is valid for a wide rage of vector based quatizatio schemes The proposed upper boud turs out to be quite tight maily due to the exact evaluatio of the expected value of matrix determiat as opposed to similar studies usig Hadamard iequality to upper boud the determiat as i [1] ad usig approximate desity fuctio of determiat expressio ad partitio cell approximatio i [11], [1] As a byproduct, a absolute upper boud to LRF MIMO capacity usig precodig based quatizatio is also herei derived by utilizig a boudig distributio for Grassmaia beamformig [13] The outlie is as follows The system model is explaied i Sectio II MIMO capacity expressios for LRF are obtaied i Sectio III A aalytical upper boud for the LRF MIMO capacity is derived i Sectio IV Two exemplary quatizatio schemes will be studied i Sectio V ad correspodig umerical results are preseted i Sectio VI The paper is cocluded with Sectio VII II SYSTEM MODEL The followig otatio is used throughout the mauscript Boldface lower ad upper-case letters deote colum vectors ad matrices, respectively Scalars are deoted by plai lower-case letters The superscript deotes the complex cojugate for scalars ad cojugate traspose for vectors ad matrices The absolute value of a scalar is show with The idetity matrix is show with I The trace operator ad determiat are deoted by tr ad, respectively The autocorrelatio matrix for a radom vector a is R a = E[aa ]
2 where E[ ] stads for the expected value operator The i, j th elemet of a matrix A is deoted by A i,j The geeral expressio for a poit-to-poit MIMO chael with r receive ateas ad t trasmit ateas is give by ỹ = H x + w, where ỹ is the received vector, H is the r t chael matrix, x is the trasmitted vector, ad w is the zero-mea circularly symmetric complex Gaussia ZMC- SCG white spatially ad temporally oise with ormalized variace 1 The chael matrix H is comprised of idepedet ZMCSCG radom variables with variace 1 Cosiderig a bloc fadig model, the chael matrix is assumed to be costat durig a coherece iterval sigificatly larger tha symbol duratio A fixed average power is allotted for each trasmissio which correspods to settig trr x P Sec 13 i [9] I the case that perfect chael iformatio is available both at the trasmitter ad receiver, sigular value decompositio SVD is applied to decompose the MIMO chael ito mi r, t parallel subchaels over which multiple streams may be trasmitted [14] The followig equivalet expressio is obtaied for the received vector whe SVD is performed to attai H = UDV : U ỹ = DV x + U w 1 The etries of D are tae to be decreasig without loss of geerality The trasmitted vector ca be writte i geeral as i x = PΛx where P is a precodig matrix, Λ is a diagoal matrix used to distribute power amog subchaels, ad x is the origial iformatio vector assumed to have R x = I mir, t If the precodig matrix is chose to be P = V, by the uitary property of the precodig matrix V V = I y = DΛx + w, where y = U ỹ ad w = U w Sice both D ad Λ are diagoal ad R w = I, the chael is decomposed ito parallel subchaels The capacity is achieved by Λ obtaied through the waterfillig procedure [14] with the costrait that trλ P We ote here that the colums of matrix V are isotropically distributed o the t dimesioal complex uit circle whe cosidered over the realizatios of H Whe there is oly a partial CSI i trasmitter due to fiite rate feedbac, oe has imperfect precodig ad power distributio matrices deoted by V f ad Λ f, respectively Eq ow becomes y = DV V f Λ f x + w 3 which suggests that subchaels ow iterfere with each other sice V V f I, i geeral We will ivestigate the capacity of LRF MIMO chaels based o 3 I order to reduce the rate of the feedbac chael, the idea of reduced precodig ca be used [15] I this scheme, the umber of beamformers used i a spatial multiplexig system is adaptively varied i order to miimize probability of symbol vector error or to maximize capacity by allocatig equal power Λ f = P I to selected subchaels [15], [11] Trasmittig oly the precodig vectors correspodig to the strogest subchaels will suffice to maximize commuicatio rate over MIMO chaels Thus, this strategy allows efficiet utilizatio of the feedbac bits by quatizig oly relevat precoders The aalytical boud for limited rate feedbac MIMO capacity to be obtaied i sectio IV ca be used to determie the umber of precoders to be used at each average SNR value i order to maximize the spectral efficiecy The idea of reduced precodig ad the utilizatio of feedbac for precoders are ot oly useful at low SNR values but also at high SNR especially for MIMO systems with t > r [] III CAPACITY WITH LIMITED RATE PRECODING Eq 3 ca be writte with a equivalet chael matrix H = DV V f as i y = HΛ f x + w 4 Notig that Λ f = P I, the capacity of this scheme which maes use of precoders is give by { C pre = E log det I + P } HH 5 ad ca be achieved with MMSE estimatio ad successive iterferece cacellatio [16] [18] where the equivalet chael matrix H ca be writte as H = DV V f d 1 d = d v 1 v v [ ] v1f v f The equivalet chael matrix H has its i, j th elemet as H i,j = d i vi v jf, where v i is the i th colum of V ad, v jf is the j th colum of V f Defiig V ij = vi v jf, oe ca evaluate HH i,j = d i d j =1 V i V j 6 7 Evaluatio of the capacity i 5 requires the probability distributios of V ij = vi v jf for i, j {1,,, } ad hece the quatizatio rule used for limited rate feedbac has to be specified A set of N f vectors {q 1,q,,q N f } geerated to costruct the quatizatio codeboo are defied where N f stads for the umber of feedbac bits per precodig vector Quatizatio vectors are legth- t complex vectors o the t - dimesioal complex uit circle ad the quatizatio while obtaiig the precodig vectors is determied by the followig rule used i LRF MIMO studies [4] [6], [19]: v if = arg max q j, j=1,, N f v i q j 8 There are two types of radom variables i 5 whose distributios ad depedece properties have to be determied i order to evaluate C pre First, the cumulative distributio fuctio cdf of V ii = v i v if is peculiar to the give quatizatio codeboo ad rule, ad we will ivestigate the cdf of V ii for two differet quatizatio methods i Sectio V Moreover, the
3 3 cdf for V ij s for i j is eeded I [4], [5], [19], the cdf of the squared absolute ier product betwee two isotropically distributed legth- t complex uit vectors is give as F t o x = 1 1 x t 1, x 1, x < 1, x > 1 The same result ad hece cdf hold for the case of oe fixed vector ad a isotropically distributed vector sice oe of them beig isotropically distributed is sufficiet for the result [4] Bearig i mid that vi v jf correspods to projectio of v i oto v jf, the followig holds by orthogoality of v i s i our problem 1 v v 1f = v v 1f vj v 1fv j 1 j=1 for =,, Defiig v 1f = v 1f 1 j=1 v j v 1fv j, the vector v 1f is i the ull space of v i s, i = 1,, 1, where the ull space has dimesio t +1 The squared orm of v 1f is v 1f v 1f = 1 1 j=1 v j v 1f = 1 1 j=1 V j1 Cosiderig the projectio of a fixed vector v oto a isotropically distributed vector v 1f which is of dimesio t +1, oe obtais the followig coditioal probability distributio fuctio for v v 1f by usig 9: F v v 1f x v i v 1f = a i, i = 1,, 1 = Fo t +1 x i=1 a i for =,, The expectatio i 5 is over the chael matrix H or equivaletly, over V ij s For a give chael realizatio, v i s are fixed ad the quatized precodig vectors v if s are chose idepedetly of each other by the rule give i 8 Hece, v if s are idepedet of each other o the coditio that v i s are give Oe should ote that V V f product ivolves V ij terms ad the phase of V ij becomes relevat i this case Recallig that the vectors are isotropically distributed, the phases of all the radom variables correspodig to the projectios of the precoders oto quatized precoders are idepedet ad uiformly distributed i [, π] sice the quatizatio rule i 8 is blid to multiplicatio of all the etries of a quatizatio vector by a complex umber α of uity amplitude as a b = αa b To summarize, it holds true that V ij has a uiformly distributed phase i, π ad it is idepedet of V i for all j ad V lj for all l i for give v i s sice owig v i oly is sufficiet to determie v if for the give rule i 8 Isotropical distributio implies that V ii = v i v if s for i = 1,, are idetically distributed ad idepedet radom variables Similarly, a correspodig distributio holds for V j = v v jf ad its cdf has a form idetical to that give i 11 IV A CAPACITY UPPER BOUND The capacity expressio i 5 is impractical to be used i practical system desig sice it eeds the distributio of 9 V ij s ad the expectatio over V ij s distributios We will obtai a very tight aalytical upper boud for capacity of the -precoder scheme ad this aalytical boud eeds oly the expectatios E 11 = E V 11 ad E 1 = E V 1 to be evaluated The expected values will be deoted with E ij = E V ij ad E ii s are the same for all i s ad ca be evaluated for a give quatizatio codeboo Whe the value of E 11 is give, the value of E 1 ca be calculated easily as follows Due to the isotropical distributio of v i s, E ij is same as E 1 for all i j The value of E 1 ca be foud i terms of E 11 by usig 11 as = E[ V 1 ] = E [ E[ V 1 V 11 ] ] = xf V1 x V 11 = af V11 adxda 1 a x t 1 x t 3 dx f V11 1 a 1 a ada = 1 a t 1 f V 11 ada = 1 E 11 t 1 1 This result i 1 is ituitive sice, after projectig v 1f o v 1, t 1 orthoormal vectors {v,v 3,,v t } are left Due to isotropical distributio, the power i the part of v 1f orthogoal to v 1 is distributed equally betwee t 1 orthoormal vectors i average The capacity for the precoder scheme ca be upper bouded by usig Jese s iequality such that E {log I + P } { HH log E I + P } HH 13 The capacity boud of the -precoder case give i 13 ca be writte i terms of E 11 ad E 1 by usig the result of the lemma give i Appedix as C pre log 1 + j 1 S j S, j j 1 =1 S P P E [ d a1d a d a ] E a1j 1 E a j j S, j j 1,,j 1 14 where E aij i = E [ V aij i ] { E11 = E [ V = 11 ] if a i = j i E 1 = E [ V 1 ], if a i j i 15 S = {1,,, }, S = {a 1,,a }; ad P is the set cotaiig all possible, combiatios of S The term j 1 S j S, j j 1,,j 1 E a 1j 1 E a j does ot deped o which elemet combiatio a 1, a,, a chose from set P is used i 14 sice j i s for i = 1,,, are chose from set S = {1,,, } Usig this fact, oe ca further simplify the capacity boud give i 14 by selectig
4 4 a 1, a,, a as 1,,, such that The secod method is the oe that maximizes the capacity { } by icreasig the projectio power, amely V P 11 By defiig C pre log 1 + E d a1 d a d a a radom variable z as z = 1 v1v 1f, we ca say that for =1 S P a good beamformer, the expected value of z has to be as close as possible to zero A upper boud for the cdf of z, F z z, E 1j1 E j 16 is give i [13] such that F z z F z z for z 1 ad j 1 S j S, j j 1 j S, j j 1,,j 1 E { S P d a1 d a d a } 1 s for = 1,, are required to evaluate the capacity boud i 16 ad oe does ot N f z t 1 1 t 1, z < F eed to calculate E [ z z = N f 1 19 d a1 d a d a ] 1 t 1 for all possible, 1, z N f combiatios a 1,, a from set S separately Istead, the A good beamformer shall try to come as close as possible to expectatio of the sum of all possible joit momets are this distributio which we refer to as the boudig distributio ecessary Hece, oly the expectatio of the sum of all possible joit momets E { S P d a1 d a d a } hereafter [13] For a hypothetical quatizatio method that has to attais this boudig distributio, we ca evaluate the E 11 be foud This ca be foud by usig Mote Carlo simulatios ad E 1 values i order to costruct a upper boud to the or aalytically by usig curretly available results for the joit capacity of limited rate feedbac MIMO scheme that oe of ordered momets i a closed form preseted i [] ad this the quatizatio methods ca exceed E 11 ca be foud as maes the proposed boud completely aalytical give below As will be see i Sectio VI, the capacity boud i 16 is very tight ad ca be used for practical purposes With the help of this aalytical boud, oe ca desig a adaptive E 11 = 1 E[z] = 1 N t 1 f 1 t t 1 t N f MIMO system that ca chage the umber of precoders ad ad E the feedbac rate accordig to the curret average SNR value 1 by 1 The value of E 11 obtaied by this boudig distributio i is the maximum value of E The boud give i 16 is valid for may vector quatizatio 11 that ca methods as log as V ij = vi v be achieved amog all quatizatio codeboos usig the jf has a distributio such that quatizatio rule i 8 for give N the phase of V ij is uiformly distributed i, π which f ad t values is idepedet of V i for all j ad V lj for all l i The quatizatio rule give i 8 is sufficiet but ot ecessary VI NUMERICAL RESULTS to meet this coditio Thus, the results of this sectio ca be applied to a wide variety of quatizatio schemes Moreover, the boud is also valid for o quatizatio cases ad we ca easily costruct a upper boud for the exact capacity of a MIMO system that uses of its strogest subchaels by simply settig E 11 = 1 ad E 1 = i 16 V BOUNDING DISTRIBUTION AND RVQ I this sectio, we will preset two exemplary quatizatio methods that will be used to produce some umerical results i sectio VI The first method is radom vector quatizatio RVQ i which a codeboo is geerated radomly ad the closest vector i the codeboo is coveyed to the trasmitter accordig to the rule give i 8 [19] The use of RVQ as a vector quatizatio scheme allows simpler aalysis ad ca be helpful i desigig practical limited rate feedbac MIMO systems The probability distributio fuctio of V 11 has bee readily obtaied i [4], [5], [19] as t F v 1 v 1f x = Fo x N f 17 The phase of V ij is idepedet of its magitude ad uiformly distributed i [, π] The expected value of a radom variable distributed with 17 is evaluated i [19] as E[ V 11 ] = E 11 = 1 N f B N f t,, 18 t 1 where Bx, y = ΓxΓy Γx+y is the Beta fuctio ad the Gamma fuctio is give by Γx = t x 1 e t dt while E 1 = E V 1 ca be calculated from 1 I this sectio, the capacity upper boud obtaied for LRF MIMO give i 16 is evaluated for RVQ ad the boudig distributio Radom variables are geerated by the iversio method [1] i simulatios ad placed i 5 to obtai ergodic capacity The radom variables V ii s are geerated first while the others are draw based o the distributio give i 11 ad each data poit is obtaied by geeratig 1, realizatios I Fig 1, 6 3 MIMO system capacity uder limited rate feedbac with N f = 8 is evaluated by usig RVQ ad the boudig distributio for ad 3-precoder schemes where the capacity boud give i 16 is also foud by usig E 11 ad E 1 values for these schemes It is see i the figure that the capacity upper boud is very tight It is 3-5 db away from the 6 3 LRF MIMO capacities It is further observed that the RVQ scheme is almost optimal sice RVQ ad boudig distributio capacities are quite close to each other There is a db differece betwee these two capacities ad hece we ca say that RVQ ca be used as a practical quatizatio techique that attai rates quite close to the capacity with tolerable N f values This also justifies its use i the literature for aalysis purposes Moreover, whe compared to the exact MIMO chael capacity obtaied with waterfillig WF that uses a short-term power costrait Sec 13 i [9], there is a 13 db loss i limited feedbac icremetal precodig scheme that uses RVQ with N f = 8 which is well predicted by the proposed boud at high SNR I Fig, RVQ is used as a quatizatio techique ad the capacities of precoder 8 4 MIMO schemes are evaluated with their correspodig bouds for N f = 1 ad =, 3,
5 5 5 Upper capacity boud usig boudig distributio with N f Upper capacity boud usig RVQ with N f Capacity simulatio usig boudig distributio with N f Upper capacity boud usig boudig distributio with N f =8 for precoders x4 MIMO capacity with WF, o quatizatio 8x4 equipower capacity Upper capacity boud usig RVQ with N f =1 for 4 precoders =1 for 4 precoders Upper capacity boud usig RVQ with N f =1 for 3 precoders Rate bps/hz 15 1 Upper capacity boud usig RVQ with N =8 for precoders f Capacity simulatio usig boudig distributio with N f =8 for precoders =8 for precoders 6x3 MIMO capacity with WF, o quatizatio Rate bps/hz 5 15 =1 for 3 precoders Upper capacity boud usig RVQ with N f =1 for precoders =1 for precoders Average SNR db Fig MIMO capacities ad upper bouds for ad 3-precoder LRF schemes RVQ ad boudig distributio are used with N f = Average SNR db Fig 8 4 MIMO capacities ad upper bouds for, 3 ad 4-precoder LRF schemes RVQ is used with N f = 1 ad 4 The equipower scheme, which correspods to capacity with o CSI at trasmitter ad allocates equal power amog the trasmit ateas, is also depicted for compariso It ca be observed that the capacity bouds evaluated with the E 11 value of RVQ for N f = 1 is quite fie ad 15-4 db away from the ergodic capacity with =, 3, ad 4 Furthermore, it is see that the -precoder capacity with N f = 1 is better tha the equipower scheme up to 6dB, 3-precoder scheme is better up to 135 db The 4-precoder scheme that uses all the degrees of freedom i the system always has higher capacity tha the equipower scheme At high SNR, there is a 16 db differece betwee the 4-precoder ad equipower schemes Whe compared to the exact capacity with WF, there is a 14 db loss i limited feedbac icremetal precodig scheme at high SNR ad this ca be compesated by icreasig N f N f = 1 may seem to be large for practical systems but this large value is due to the large trasmit atea umber t = 8 I cotrast, a reasoable N f value is sufficiet to reach the capacity for small size MIMO system I case of limited feedbac, the losses are approximately equal to the SNR loss due to quatizatio 1 log 1 1 E 11 db Ituitively speaig, the istataeous effective SNR of the chael for sigle precoder case is P V 11 ad hece there occurs a 1 log 1 E 11 db SNR loss i the LRF sceario This holds approximately true for the geeral precoder scheme As N f icreases, E 11 asymptotically becomes 1 as observed i 18 ad so that the capacity of LRF MIMO approaches to the capacity with o quatizatio The capacity upper boud i 16 is maximized at E 11 = 1 ad E 1 = so a good quatizatio scheme should have a E 11 value which is as close to 1 as possible for a give umber of feedbac bits Amog the give vector quatizatio techiques ad N f, the best quatizatio codeboo is the oe that gives the highest E 11 value ad thus, it has the greatest capacity boud value i 16 As a result, we ca use the capacity boud give i 16 to evaluate the performace of differet quatizatio schemes For a give quatizatio scheme, it ca be used to determie the umber of precoders to be used at each average SNR value I other words, for a give MIMO system ad the quatizatio techique with N f value, oe calculates the E 11 value of the quatizatio ad the expectatio of the sum of possible joit momets E { S P d a1 d a d a } i 16 oly oce After that, these two values ca be used to costruct the boud i 16 easily With the help of this boud, it is possible to determie the SNR regios i which the umber of precoders to be used to maximize capacity are specified durig the operatio of the studied icremetal precodig scheme For our exemplary quatizatio i Fig, it is see that usig precoders up to db, usig 3 precoders betwee ad 5 db, ad usig 4 precoders above 5 db maximizes the capacity ad this strategy always results i sigificatly higher capacity tha the equipower scheme especially for t > r The mai coclusio is that the tightess of the proposed boud is established The upper boud ca be applied to may LRF MIMO schemes ad also to other problems where the exact evaluatio of the determiat expected value is eeded VII CONCLUSION We developed a tight upper boud to poit-to-poit LRF MIMO capacity that is valid for a large body of vector quatizatio schemes The umber of precoders used i a practical system ca be determied for each average SNR value based o the upper boud developed i this paper We furthermore evaluated the upper boud usig a boudig distributio from Grassmaia beamformig which resulted i the observatio that the simple RVQ techique performs quite close to capacity upper boud Perfect chael estimatio at the receiver is assumed i this study Future studies will iclude the cosideratio of imperfect chael estimatio ad its delayed trasmissio to the trasmitter side withi a limited rate feedbac sceario i a mobile system Also, practical quatizatio methods for MIMO systems will be ivestigated withi the framewor developed i this paper
6 6 APPENDIX: PROOF OF 14 Defie a matrix B = P HH with elemets B i,j = P d id j =1 V ivj ad aother matrix A = I + B For the capacity boud of the precoder MIMO scheme, the determiat of the matrix A is ecessary i evaluatig 14 ad it ca be foud by Leibiz formula []: deta = sgσ, 1 Σ S, =1,,! i=1 A i,σ i where σ i is the i th elemet of Σ which is the th elemet of the permutatio group S, ad S icludes all possible permutatios of the set S = {1,,, } There are! differet permutatios of S ad hece S is composed of! permutatios The fuctio sg of permutatios i the permutatio group S returs +1 or 1 for eve ad odd permutatios, respectively [3] Recallig that A ij equals B ij for i j ad 1 + B ij otherwise, oe ca write the determiat expressio i a compact form i terms of B ij s directly from 1 after a careful ispectio as deta = 1 + Σ l S, l=1,,! i=1 =1 S =a 1,,a P B ai,σ isg Σ l l, where P is the set cotaiig all possible, combiatios of {1,,, } S icludes all possible permutatios of S = {a 1, a,,a } ad there are! differet permutatios i S I the above expressio, the set of elemet combiatio from S = {1,,, } is determied first Cosiderig the Σ a1,,a P term i the summatio i, σ l i is the i th elemet of Σ l which is the l th elemet of the permutatio group S Recallig that B ij = P d id j =1 V iv j, oe ca put B ij ito deta expressio give i ad obtai the followig: P d a i d σ l i deta = 1 + =1 S P Σ l S i=1 V aimv σ l im m=1 sg Σ l 3 The above expressio ca be simplified as follows Defiig S = {a 1,, a } ad a partitio S1,S,, Sp S which are disjoit sets that satisfy S1 S Sp = S = {a 1,,a }, s i j is the jth elemet i Si so that s i j, j = 1,, Si, are the elemets belogig to S i where S i is the cardiality of Si for i = 1,,,p 1 p Although there are may terms i 3, the oly terms that have ozero mea are the oes all composed of squared forms V ij This is due to the reaso that V ij has a uiformly distributed phase i, π which is idepedet of V i for all j ad V lj for all l i This uiform phase distributio will result i a zero expected value for ay term which has a o-squared form of V ij Usig this fact ad after a few straightforward steps, the followig simplified expressio ca be obtaied from 3 after taig expectatio P E{detA} = E{1 + d a1d a d a =1 S P, t=1,,p S1,S,,S p [S ] p=1,, Σ l t S t S p i V s i jz i sg Σ l }, i=1 z i=1 j=1 4 where z 1 z z p, Σ l = Σ l 1,, Σ l p, ad St is the permutatio group that icludes all possible permutatios of the elemets i St I the above equatio, [S ] p=1,, is the set that icludes all possible p elemet partitios S 1, S,, S p of the set S for p = 1,, Therefore, the summatio S1,S,,S p [S ] p=1,, Σ l t S t, t=1,,p i 4 is over all differet permutatios of all possible S1, S,,S p partitios, ie, the summatio shows that p elemet partitio is chose from [S ] p=1,, first ad the, the secod summatio Σ l t S t, t=1,,p is tae over all differet permutatios of the chose partitio of S Actually, the summatio Σ l t S t, t=1,,p is the shorthad otatio of Σ l 1 S 1 Σ l S Σ l p S p ad sg Σ l taes values +1 or 1 depedig o whether Σ l = Σ l 1,, Σ l p is a eve or odd permutatio Lemma 1: The expressio give i 4 ca be simplified as E{detA} = E{1 + j 1 S j S, j j 1 =1 S P j S, j j 1,,j 1 P d a1d a d a V a1j 1 V a j } 5 Proof: The lemma suggests that the oly remaiig terms i 4 are the terms resultig from the partitio of S1,,S p with p = The oly possible partitio is the S1 = {a 1 }, S = {a },,S = {a }, ad Si = 1 for all i s Eq 4 reduces to 5 for p = ad z 1 z z Ay term icluded withi the summatio S1,S,,S p [S ] p=1,, 4 but ot i 5 ca also be writte as p S i i=1 j=1 V s i jz i Σ l t S t, t=1,,p i 6 for ay give z 1, z,, z p betwee 1 ad where z 1 z z p the terms with the same z i s are already i aother partitio of S s i j is the jth elemet of Si as defied before For the above term, at least oe of the Si s has cardiality Si greater or equal to for some i sice p <,
7 7 ie, max 1 i p Si Note that the terms i 5 have Si = 1 for i = 1,,,p ad p = The terms i the form give i 6 origiate from the term give below i 4 S p i E V s i jz i sg Σ l 7 i=1 z i=1 j=1 Now focus o the summatio i 4 over permutatios Σ l t S t for t = 1,,p Fix z 1,,z p z 1 z z p madated by 6 ad all the Σ l 1, Σ l,,σ l p permutatios except for oe of the Σ l c with S c The, 6 ca be writte as S p i S c V s i jz i V s c jz c sg Σ l 8 i c j=1 j=1 alog with the permutatio sig By taig the summatio over differet Σ l c permutatios, oe ca get sg Σ l t, t = 1,,p ad t c Σ l c S c p S i i c j=1 V s i jz i S c V s c jz c sg Σ l c =, 9 j=1 sice there are equal umber of eve ad odd permutatios of Sc set if S c with opposite sigs [3] I other words, i the above equatio sgσ l c is +1 for S c! times ad 1 for S c! times agai Therefore, the expressio i 9 goes to zero ad this ca be doe for other Σ l 1,, Σ l p permutatios The terms preseted i 6 cacel each other i 4 ad 4 reduces to the equatio give i 5 QED The simplified form of the capacity boud i 5 is importat, sice E [ V a1j 1 V a j ] = E [ V a1j 1 ] E [ V a j ] 3 by the idepedece of V a1j 1,, V a j Sice the distributio of V aij i s are idetical ad same as the distributio of V 11 if a i = j i ad equal to the distributio of V 1 if a i j i, oe eed oly the expected values E [ V 11 ] = E 11 ad E [ V 1 ] = E 1 i order to calculate the expressio give i 5 [6] N Jidal, MIMO broadcast chaels with fiite rate feedbac, Proc IEEE Globecom, 5 [7] I E Telatar, Capacity of multi-atea Gaussia chaels, Europ Tras Telecommu, vol 1, pp , Nov/Dec 1999 [8] Z Zhou, B Vucetic, M Dohler, ad Y Li, MIMO systems with adaptive modulatio, IEEE Tras Vehicular Tech, vol 54, o 5, pp , Sept 5 [9] A Goldsmith, Wireless Commuicatios, Cambridge Uiversity Press, 5 [1] Amir D Dabbagh ad David J Love, Feedbac rate-capacity loss tradeoff for limited feedbac MIMO systems, IEEE Trasactios o Iformatio Theory, vol 5, o 5, pp 19, May 6 [11] Jue Chul Roh ad Bhasar D Rao, Desig ad aalysis of MIMO spatial multiplexig systems with quatized feedbac, IEEE Trasactios o Sigal Processig, vol 54, o 8, pp , August 6 [1] Wiroosa Satipach ad Michael L Hoig, Asymptotic performace of MIMO wireless chaels with limited feedbac, Military Commuicatios Coferece, 3 MILCOM 3 IEEE, vol 1, pp , October 3 [13] B Giaais S Zhou, Z Wag, Quatifyig the power loss whe trasmit beamformig relies o fiite-rate feedbac, IEEE Tras o Wireless Commuicatios, vol 4, o 4, pp , July 5 [14] EG Larsso ad P Stoica, Space-Time Bloc Codig for Wireless Commuicatios, Cambridge Uiversity Press, 3 [15] David J Love ad Robert W Heath Jr, Multi-mode precodig usig liear receivers for limited feedbac MIMO systems, 4 IEEE Iteratioal Coferece o Commuicatios, vol 1, pp , Jue 4 [16] GD Forey Jr, O the role of MMSE estimatio i approachig the iformatio-theoretic limits of liear Gaussia chaels: Shao meets Wieer, Proc 3 Allerto Cof, pp , Oct 3 [17] GD Forey Jr, Shao meets Wieer II: O MMSE estimatio i successive decodig schemes, Proc 4 Allerto Cof, 4 [18] P Stoica, Y Jiag, ad J Li, O MIMO chael capacity: A ituitive discussio, IEEE Sigal Processig Mag, pp 83 84, May 5 [19] C Au-Yeug ad DJ Love, O the performace of radom vector quatizatio limited feedbac beamformig i a MISO system, IEEE Tras Wireless Commuicatios, vol 6, o, pp , February 7 [] Si Ji, Matthew R McKay, Xiqi Gao, ad Iai B Colligs, MIMO multichael beamformig: Ser ad outage usig ew eigevalue distributios of complex ocetral wishart matrices, IEEE Trasactios o Commuicatio, vol 56, o 3, pp , March 8 [1] L Devroye, No-Uiform Radom Variate Geeratio, New Yor: Spriger-Verlag, 1986 [] H Campbell, Liear Algebra With Applicatios, Appleto Cetury Crofts, 1971 [3] Joh D Dixo ad Bria Mortimer, Permutatio Groups, Graduate Texts i Mathematics, Spriger-Verlag, 1996 REFERENCES [1] T Yoo, E Yoo, ad A Goldmisth, MIMO capacity with chael ucertaity: Does feedbac help?, Proc IEEE Globecom, pp 96 1, 4 [] E Biglieri, R Calderba, A Costatiides, A Goldsmith, A Paulraj, ad H Vicet Poor, MIMO Wireless Commuicatios, Cambridge Uiversity Press, 7 [3] JC Roh ad BD Rao, Chael feedbac quatizatio methods for MISO ad MIMO systems, Proc IEEE PMIRC, vol, pp 85 89, Sept 4 [4] DJ Love, RW Heath Jr, ad T Strohmer, Grassmaia beamformig for multiple-iput multiple-output wireless systems, IEEE Tras Iform Theory, vol 49, o 1, pp , Oct 3 [5] KK Muavilli, A Sabharwal, E Erip, ad B Aazhag, O beamformig with fiite rate feedbac i multiple-atea systems, IEEE Tras Iform Theory, vol 49, o 1, pp , Oct 3
Lecture 7: MIMO Architectures Theoretical Foundations of Wireless Communications 1. Overview. CommTh/EES/KTH
: Theoretical Foudatios of Wireless Commuicatios 1 Thursday, May 19, 2016 12:30-15:30, Coferece Room SIP 1 Textbook: D. Tse ad P. Viswaath, Fudametals of Wireless Commuicatio 1 / 1 Overview Lecture 6:
More informationOPTIMAL PIECEWISE UNIFORM VECTOR QUANTIZATION OF THE MEMORYLESS LAPLACIAN SOURCE
Joural of ELECTRICAL EGIEERIG, VOL. 56, O. 7-8, 2005, 200 204 OPTIMAL PIECEWISE UIFORM VECTOR QUATIZATIO OF THE MEMORYLESS LAPLACIA SOURCE Zora H. Perić Veljo Lj. Staović Alesadra Z. Jovaović Srdja M.
More informationOFDM Precoder for Minimizing BER Upper Bound of MLD under Imperfect CSI
MIMO-OFDM OFDM Precoder for Miimizig BER Upper Boud of MLD uder Imperfect CSI MCRG Joit Semiar Jue the th 008 Previously preseted at ICC 008 Beijig o May the st 008 Boosar Pitakdumrogkija Kazuhiko Fukawa
More informationProblem Set 4 Due Oct, 12
EE226: Radom Processes i Systems Lecturer: Jea C. Walrad Problem Set 4 Due Oct, 12 Fall 06 GSI: Assae Gueye This problem set essetially reviews detectio theory ad hypothesis testig ad some basic otios
More informationThe Maximum-Likelihood Decoding Performance of Error-Correcting Codes
The Maximum-Lielihood Decodig Performace of Error-Correctig Codes Hery D. Pfister ECE Departmet Texas A&M Uiversity August 27th, 2007 (rev. 0) November 2st, 203 (rev. ) Performace of Codes. Notatio X,
More informationSession 5. (1) Principal component analysis and Karhunen-Loève transformation
200 Autum semester Patter Iformatio Processig Topic 2 Image compressio by orthogoal trasformatio Sessio 5 () Pricipal compoet aalysis ad Karhue-Loève trasformatio Topic 2 of this course explais the image
More informationEconomics 241B Relation to Method of Moments and Maximum Likelihood OLSE as a Maximum Likelihood Estimator
Ecoomics 24B Relatio to Method of Momets ad Maximum Likelihood OLSE as a Maximum Likelihood Estimator Uder Assumptio 5 we have speci ed the distributio of the error, so we ca estimate the model parameters
More information1 Introduction to reducing variance in Monte Carlo simulations
Copyright c 010 by Karl Sigma 1 Itroductio to reducig variace i Mote Carlo simulatios 11 Review of cofidece itervals for estimatig a mea I statistics, we estimate a ukow mea µ = E(X) of a distributio by
More informationTopic 9: Sampling Distributions of Estimators
Topic 9: Samplig Distributios of Estimators Course 003, 2016 Page 0 Samplig distributios of estimators Sice our estimators are statistics (particular fuctios of radom variables), their distributio ca be
More informationSymmetric Two-User Gaussian Interference Channel with Common Messages
Symmetric Two-User Gaussia Iterferece Chael with Commo Messages Qua Geg CSL ad Dept. of ECE UIUC, IL 680 Email: geg5@illiois.edu Tie Liu Dept. of Electrical ad Computer Egieerig Texas A&M Uiversity, TX
More informationEECS564 Estimation, Filtering, and Detection Hwk 2 Solns. Winter p θ (z) = (2θz + 1 θ), 0 z 1
EECS564 Estimatio, Filterig, ad Detectio Hwk 2 Sols. Witer 25 4. Let Z be a sigle observatio havig desity fuctio where. p (z) = (2z + ), z (a) Assumig that is a oradom parameter, fid ad plot the maximum
More informationLecture 2: Monte Carlo Simulation
STAT/Q SCI 43: Itroductio to Resamplig ethods Sprig 27 Istructor: Ye-Chi Che Lecture 2: ote Carlo Simulatio 2 ote Carlo Itegratio Assume we wat to evaluate the followig itegratio: e x3 dx What ca we do?
More informationGeometry of LS. LECTURE 3 GEOMETRY OF LS, PROPERTIES OF σ 2, PARTITIONED REGRESSION, GOODNESS OF FIT
OCTOBER 7, 2016 LECTURE 3 GEOMETRY OF LS, PROPERTIES OF σ 2, PARTITIONED REGRESSION, GOODNESS OF FIT Geometry of LS We ca thik of y ad the colums of X as members of the -dimesioal Euclidea space R Oe ca
More informationThe Fading Number of Multiple-Input Multiple-Output Fading Channels with Memory
The Fadig Number of Multiple-Iput Multiple-Output Fadig Chaels with Memory Stefa M. Moser Departmet of Commuicatio Egieerig Natioal Chiao Tug Uiversity NCTU Hsichu, Taiwa Email: stefa.moser@ieee.org Abstract
More informationProbability of error for LDPC OC with one co-channel Interferer over i.i.d Rayleigh Fading
IOSR Joural of Electroics ad Commuicatio Egieerig (IOSR-JECE) e-issn: 2278-2834,p- ISSN: 2278-8735.Volume 9, Issue 4, Ver. III (Jul - Aug. 24), PP 59-63 Probability of error for LDPC OC with oe co-chael
More informationCEE 522 Autumn Uncertainty Concepts for Geotechnical Engineering
CEE 5 Autum 005 Ucertaity Cocepts for Geotechical Egieerig Basic Termiology Set A set is a collectio of (mutually exclusive) objects or evets. The sample space is the (collectively exhaustive) collectio
More informationBER results for a narrowband multiuser receiver based on successive subtraction for M-PSK modulated signals
results for a arrowbad multiuser receiver based o successive subtractio for M-PSK modulated sigals Gerard J.M. Jasse Telecomm. ad Traffic-Cotrol Systems Group Dept. of Iformatio Techology ad Systems Delft
More informationOptimally Sparse SVMs
A. Proof of Lemma 3. We here prove a lower boud o the umber of support vectors to achieve geeralizatio bouds of the form which we cosider. Importatly, this result holds ot oly for liear classifiers, but
More informationLecture 27. Capacity of additive Gaussian noise channel and the sphere packing bound
Lecture 7 Ageda for the lecture Gaussia chael with average power costraits Capacity of additive Gaussia oise chael ad the sphere packig boud 7. Additive Gaussia oise chael Up to this poit, we have bee
More informationIt is always the case that unions, intersections, complements, and set differences are preserved by the inverse image of a function.
MATH 532 Measurable Fuctios Dr. Neal, WKU Throughout, let ( X, F, µ) be a measure space ad let (!, F, P ) deote the special case of a probability space. We shall ow begi to study real-valued fuctios defied
More informationAlgebra of Least Squares
October 19, 2018 Algebra of Least Squares Geometry of Least Squares Recall that out data is like a table [Y X] where Y collects observatios o the depedet variable Y ad X collects observatios o the k-dimesioal
More informationLinear regression. Daniel Hsu (COMS 4771) (y i x T i β)2 2πσ. 2 2σ 2. 1 n. (x T i β y i ) 2. 1 ˆβ arg min. β R n d
Liear regressio Daiel Hsu (COMS 477) Maximum likelihood estimatio Oe of the simplest liear regressio models is the followig: (X, Y ),..., (X, Y ), (X, Y ) are iid radom pairs takig values i R d R, ad Y
More informationOrthogonal Gaussian Filters for Signal Processing
Orthogoal Gaussia Filters for Sigal Processig Mark Mackezie ad Kiet Tieu Mechaical Egieerig Uiversity of Wollogog.S.W. Australia Abstract A Gaussia filter usig the Hermite orthoormal series of fuctios
More informationResampling Methods. X (1/2), i.e., Pr (X i m) = 1/2. We order the data: X (1) X (2) X (n). Define the sample median: ( n.
Jauary 1, 2019 Resamplig Methods Motivatio We have so may estimators with the property θ θ d N 0, σ 2 We ca also write θ a N θ, σ 2 /, where a meas approximately distributed as Oce we have a cosistet estimator
More information3. Z Transform. Recall that the Fourier transform (FT) of a DT signal xn [ ] is ( ) [ ] = In order for the FT to exist in the finite magnitude sense,
3. Z Trasform Referece: Etire Chapter 3 of text. Recall that the Fourier trasform (FT) of a DT sigal x [ ] is ω ( ) [ ] X e = j jω k = xe I order for the FT to exist i the fiite magitude sese, S = x [
More informationLinear Algebra Issues in Wireless Communications
Rome-Moscow school of Matrix Methods ad Applied Liear Algebra August 0 September 18, 016 Liear Algebra Issues i Wireless Commuicatios Russia Research Ceter [vladimir.lyashev@huawei.com] About me ead of
More informationReliability and Queueing
Copyright 999 Uiversity of Califoria Reliability ad Queueig by David G. Messerschmitt Supplemetary sectio for Uderstadig Networked Applicatios: A First Course, Morga Kaufma, 999. Copyright otice: Permissio
More informationHomework Set #3 - Solutions
EE 15 - Applicatios of Covex Optimizatio i Sigal Processig ad Commuicatios Dr. Adre Tkaceko JPL Third Term 11-1 Homework Set #3 - Solutios 1. a) Note that x is closer to x tha to x l i the Euclidea orm
More informationInformation Theory Tutorial Communication over Channels with memory. Chi Zhang Department of Electrical Engineering University of Notre Dame
Iformatio Theory Tutorial Commuicatio over Chaels with memory Chi Zhag Departmet of Electrical Egieerig Uiversity of Notre Dame Abstract A geeral capacity formula C = sup I(; Y ), which is correct for
More informationAsymptotic Coupling and Its Applications in Information Theory
Asymptotic Couplig ad Its Applicatios i Iformatio Theory Vicet Y. F. Ta Joit Work with Lei Yu Departmet of Electrical ad Computer Egieerig, Departmet of Mathematics, Natioal Uiversity of Sigapore IMS-APRM
More informationLINEAR REGRESSION ANALYSIS. MODULE IX Lecture Multicollinearity
LINEAR REGRESSION ANALYSIS MODULE IX Lecture - 9 Multicolliearity Dr Shalabh Departmet of Mathematics ad Statistics Idia Istitute of Techology Kapur Multicolliearity diagostics A importat questio that
More informationStochastic Matrices in a Finite Field
Stochastic Matrices i a Fiite Field Abstract: I this project we will explore the properties of stochastic matrices i both the real ad the fiite fields. We first explore what properties 2 2 stochastic matrices
More informationMachine Learning for Data Science (CS 4786)
Machie Learig for Data Sciece CS 4786) Lecture & 3: Pricipal Compoet Aalysis The text i black outlies high level ideas. The text i blue provides simple mathematical details to derive or get to the algorithm
More informationCS284A: Representations and Algorithms in Molecular Biology
CS284A: Represetatios ad Algorithms i Molecular Biology Scribe Notes o Lectures 3 & 4: Motif Discovery via Eumeratio & Motif Represetatio Usig Positio Weight Matrix Joshua Gervi Based o presetatios by
More informationThe z-transform. 7.1 Introduction. 7.2 The z-transform Derivation of the z-transform: x[n] = z n LTI system, h[n] z = re j
The -Trasform 7. Itroductio Geeralie the complex siusoidal represetatio offered by DTFT to a represetatio of complex expoetial sigals. Obtai more geeral characteristics for discrete-time LTI systems. 7.
More informationOn Random Line Segments in the Unit Square
O Radom Lie Segmets i the Uit Square Thomas A. Courtade Departmet of Electrical Egieerig Uiversity of Califoria Los Ageles, Califoria 90095 Email: tacourta@ee.ucla.edu I. INTRODUCTION Let Q = [0, 1] [0,
More informationRandom Variables, Sampling and Estimation
Chapter 1 Radom Variables, Samplig ad Estimatio 1.1 Itroductio This chapter will cover the most importat basic statistical theory you eed i order to uderstad the ecoometric material that will be comig
More informationConfidence interval for the two-parameter exponentiated Gumbel distribution based on record values
Iteratioal Joural of Applied Operatioal Research Vol. 4 No. 1 pp. 61-68 Witer 2014 Joural homepage: www.ijorlu.ir Cofidece iterval for the two-parameter expoetiated Gumbel distributio based o record values
More informationLecture 7: Channel coding theorem for discrete-time continuous memoryless channel
Lecture 7: Chael codig theorem for discrete-time cotiuous memoryless chael Lectured by Dr. Saif K. Mohammed Scribed by Mirsad Čirkić Iformatio Theory for Wireless Commuicatio ITWC Sprig 202 Let us first
More informationAxioms of Measure Theory
MATH 532 Axioms of Measure Theory Dr. Neal, WKU I. The Space Throughout the course, we shall let X deote a geeric o-empty set. I geeral, we shall ot assume that ay algebraic structure exists o X so that
More informationc 2006 Society for Industrial and Applied Mathematics
SIAM J. MATRIX ANAL. APPL. Vol. 7, No. 3, pp. 851 860 c 006 Society for Idustrial ad Applied Mathematics EXTREMAL EIGENVALUES OF REAL SYMMETRIC MATRICES WITH ENTRIES IN AN INTERVAL XINGZHI ZHAN Abstract.
More informationPAijpam.eu ON TENSOR PRODUCT DECOMPOSITION
Iteratioal Joural of Pure ad Applied Mathematics Volume 103 No 3 2015, 537-545 ISSN: 1311-8080 (prited versio); ISSN: 1314-3395 (o-lie versio) url: http://wwwijpameu doi: http://dxdoiorg/1012732/ijpamv103i314
More information5.1 Review of Singular Value Decomposition (SVD)
MGMT 69000: Topics i High-dimesioal Data Aalysis Falll 06 Lecture 5: Spectral Clusterig: Overview (cotd) ad Aalysis Lecturer: Jiamig Xu Scribe: Adarsh Barik, Taotao He, September 3, 06 Outlie Review of
More informationInformation-based Feature Selection
Iformatio-based Feature Selectio Farza Faria, Abbas Kazeroui, Afshi Babveyh Email: {faria,abbask,afshib}@staford.edu 1 Itroductio Feature selectio is a topic of great iterest i applicatios dealig with
More informationProperties and Hypothesis Testing
Chapter 3 Properties ad Hypothesis Testig 3.1 Types of data The regressio techiques developed i previous chapters ca be applied to three differet kids of data. 1. Cross-sectioal data. 2. Time series data.
More informationEfficient GMM LECTURE 12 GMM II
DECEMBER 1 010 LECTURE 1 II Efficiet The estimator depeds o the choice of the weight matrix A. The efficiet estimator is the oe that has the smallest asymptotic variace amog all estimators defied by differet
More informationMonte Carlo Integration
Mote Carlo Itegratio I these otes we first review basic umerical itegratio methods (usig Riema approximatio ad the trapezoidal rule) ad their limitatios for evaluatig multidimesioal itegrals. Next we itroduce
More information1 of 7 7/16/2009 6:06 AM Virtual Laboratories > 6. Radom Samples > 1 2 3 4 5 6 7 6. Order Statistics Defiitios Suppose agai that we have a basic radom experimet, ad that X is a real-valued radom variable
More informationInfinite Sequences and Series
Chapter 6 Ifiite Sequeces ad Series 6.1 Ifiite Sequeces 6.1.1 Elemetary Cocepts Simply speakig, a sequece is a ordered list of umbers writte: {a 1, a 2, a 3,...a, a +1,...} where the elemets a i represet
More informationTopic 9: Sampling Distributions of Estimators
Topic 9: Samplig Distributios of Estimators Course 003, 2018 Page 0 Samplig distributios of estimators Sice our estimators are statistics (particular fuctios of radom variables), their distributio ca be
More informationMASSACHUSETTS INSTITUTE OF TECHNOLOGY 6.265/15.070J Fall 2013 Lecture 21 11/27/2013
MASSACHUSETTS INSTITUTE OF TECHNOLOGY 6.265/15.070J Fall 2013 Lecture 21 11/27/2013 Fuctioal Law of Large Numbers. Costructio of the Wieer Measure Cotet. 1. Additioal techical results o weak covergece
More informationRun-length & Entropy Coding. Redundancy Removal. Sampling. Quantization. Perform inverse operations at the receiver EEE
Geeral e Image Coder Structure Motio Video (s 1,s 2,t) or (s 1,s 2 ) Natural Image Samplig A form of data compressio; usually lossless, but ca be lossy Redudacy Removal Lossless compressio: predictive
More informationa for a 1 1 matrix. a b a b 2 2 matrix: We define det ad bc 3 3 matrix: We define a a a a a a a a a a a a a a a a a a
Math S-b Lecture # Notes This wee is all about determiats We ll discuss how to defie them, how to calculate them, lear the allimportat property ow as multiliearity, ad show that a square matrix A is ivertible
More informationTopic 9: Sampling Distributions of Estimators
Topic 9: Samplig Distributios of Estimators Course 003, 2018 Page 0 Samplig distributios of estimators Sice our estimators are statistics (particular fuctios of radom variables), their distributio ca be
More informationAn Introduction to Randomized Algorithms
A Itroductio to Radomized Algorithms The focus of this lecture is to study a radomized algorithm for quick sort, aalyze it usig probabilistic recurrece relatios, ad also provide more geeral tools for aalysis
More informationInequalities for Entropies of Sets of Subsets of Random Variables
Iequalities for Etropies of Sets of Subsets of Radom Variables Chao Tia AT&T Labs-Research Florham Par, NJ 0792, USA. tia@research.att.com Abstract Ha s iequality o the etropy rates of subsets of radom
More informationA statistical method to determine sample size to estimate characteristic value of soil parameters
A statistical method to determie sample size to estimate characteristic value of soil parameters Y. Hojo, B. Setiawa 2 ad M. Suzuki 3 Abstract Sample size is a importat factor to be cosidered i determiig
More informationModule 5 EMBEDDED WAVELET CODING. Version 2 ECE IIT, Kharagpur
Module 5 EMBEDDED WAVELET CODING Versio ECE IIT, Kharagpur Lesso 4 SPIHT algorithm Versio ECE IIT, Kharagpur Istructioal Objectives At the ed of this lesso, the studets should be able to:. State the limitatios
More informationInverse Matrix. A meaning that matrix B is an inverse of matrix A.
Iverse Matrix Two square matrices A ad B of dimesios are called iverses to oe aother if the followig holds, AB BA I (11) The otio is dual but we ofte write 1 B A meaig that matrix B is a iverse of matrix
More informationVector Quantization: a Limiting Case of EM
. Itroductio & defiitios Assume that you are give a data set X = { x j }, j { 2,,, }, of d -dimesioal vectors. The vector quatizatio (VQ) problem requires that we fid a set of prototype vectors Z = { z
More informationTR/46 OCTOBER THE ZEROS OF PARTIAL SUMS OF A MACLAURIN EXPANSION A. TALBOT
TR/46 OCTOBER 974 THE ZEROS OF PARTIAL SUMS OF A MACLAURIN EXPANSION by A. TALBOT .. Itroductio. A problem i approximatio theory o which I have recetly worked [] required for its solutio a proof that the
More informationDiscrete-Time Systems, LTI Systems, and Discrete-Time Convolution
EEL5: Discrete-Time Sigals ad Systems. Itroductio I this set of otes, we begi our mathematical treatmet of discrete-time s. As show i Figure, a discrete-time operates or trasforms some iput sequece x [
More information6.3 Testing Series With Positive Terms
6.3. TESTING SERIES WITH POSITIVE TERMS 307 6.3 Testig Series With Positive Terms 6.3. Review of what is kow up to ow I theory, testig a series a i for covergece amouts to fidig the i= sequece of partial
More informationRandom Matrices with Blocks of Intermediate Scale Strongly Correlated Band Matrices
Radom Matrices with Blocks of Itermediate Scale Strogly Correlated Bad Matrices Jiayi Tog Advisor: Dr. Todd Kemp May 30, 07 Departmet of Mathematics Uiversity of Califoria, Sa Diego Cotets Itroductio Notatio
More informationOn Nonsingularity of Saddle Point Matrices. with Vectors of Ones
Iteratioal Joural of Algebra, Vol. 2, 2008, o. 4, 197-204 O Nosigularity of Saddle Poit Matrices with Vectors of Oes Tadeusz Ostrowski Istitute of Maagemet The State Vocatioal Uiversity -400 Gorzów, Polad
More informationDepartment of Mathematics
Departmet of Mathematics Ma 3/103 KC Border Itroductio to Probability ad Statistics Witer 2017 Lecture 19: Estimatio II Relevat textbook passages: Larse Marx [1]: Sectios 5.2 5.7 19.1 The method of momets
More informationEXAMINATIONS OF THE ROYAL STATISTICAL SOCIETY
EXAMINATIONS OF THE ROYAL STATISTICAL SOCIETY GRADUATE DIPLOMA, 016 MODULE : Statistical Iferece Time allowed: Three hours Cadidates should aswer FIVE questios. All questios carry equal marks. The umber
More informationw (1) ˆx w (1) x (1) /ρ and w (2) ˆx w (2) x (2) /ρ.
2 5. Weighted umber of late jobs 5.1. Release dates ad due dates: maximimizig the weight of o-time jobs Oce we add release dates, miimizig the umber of late jobs becomes a sigificatly harder problem. For
More informationBasics of Probability Theory (for Theory of Computation courses)
Basics of Probability Theory (for Theory of Computatio courses) Oded Goldreich Departmet of Computer Sciece Weizma Istitute of Sciece Rehovot, Israel. oded.goldreich@weizma.ac.il November 24, 2008 Preface.
More informationLast time: Moments of the Poisson distribution from its generating function. Example: Using telescope to measure intensity of an object
6.3 Stochastic Estimatio ad Cotrol, Fall 004 Lecture 7 Last time: Momets of the Poisso distributio from its geeratig fuctio. Gs () e dg µ e ds dg µ ( s) µ ( s) µ ( s) µ e ds dg X µ ds X s dg dg + ds ds
More informationChapter 9: Numerical Differentiation
178 Chapter 9: Numerical Differetiatio Numerical Differetiatio Formulatio of equatios for physical problems ofte ivolve derivatives (rate-of-chage quatities, such as velocity ad acceleratio). Numerical
More informationLinear Regression Demystified
Liear Regressio Demystified Liear regressio is a importat subject i statistics. I elemetary statistics courses, formulae related to liear regressio are ofte stated without derivatio. This ote iteds to
More informationLecture 12: November 13, 2018
Mathematical Toolkit Autum 2018 Lecturer: Madhur Tulsiai Lecture 12: November 13, 2018 1 Radomized polyomial idetity testig We will use our kowledge of coditioal probability to prove the followig lemma,
More informationApply change-of-basis formula to rewrite x as a linear combination of eigenvectors v j.
Eigevalue-Eigevector Istructor: Nam Su Wag eigemcd Ay vector i real Euclidea space of dimesio ca be uiquely epressed as a liear combiatio of liearly idepedet vectors (ie, basis) g j, j,,, α g α g α g α
More informationA Hadamard-type lower bound for symmetric diagonally dominant positive matrices
A Hadamard-type lower boud for symmetric diagoally domiat positive matrices Christopher J. Hillar, Adre Wibisoo Uiversity of Califoria, Berkeley Jauary 7, 205 Abstract We prove a ew lower-boud form of
More informationStochastic Simulation
Stochastic Simulatio 1 Itroductio Readig Assigmet: Read Chapter 1 of text. We shall itroduce may of the key issues to be discussed i this course via a couple of model problems. Model Problem 1 (Jackso
More informationa for a 1 1 matrix. a b a b 2 2 matrix: We define det ad bc 3 3 matrix: We define a a a a a a a a a a a a a a a a a a
Math E-2b Lecture #8 Notes This week is all about determiats. We ll discuss how to defie them, how to calculate them, lear the allimportat property kow as multiliearity, ad show that a square matrix A
More informationA new scalable decoder for linear space-time block codes with intersymbol interference
A ew scalale der for liear space-time lock s with itersymol iterferece Marc Kuh, Armi Wittee Uiversity of Saarlad, Istitute of Digital Commuicatios, D 6604 Saarruecke, Germay Email: marc.kuh@lnt.ui-saarlad.de,
More informationStatistical Analysis on Uncertainty for Autocorrelated Measurements and its Applications to Key Comparisons
Statistical Aalysis o Ucertaity for Autocorrelated Measuremets ad its Applicatios to Key Comparisos Nie Fa Zhag Natioal Istitute of Stadards ad Techology Gaithersburg, MD 0899, USA Outlies. Itroductio.
More informationConvergence of random variables. (telegram style notes) P.J.C. Spreij
Covergece of radom variables (telegram style otes).j.c. Spreij this versio: September 6, 2005 Itroductio As we kow, radom variables are by defiitio measurable fuctios o some uderlyig measurable space
More informationChapter 6 Principles of Data Reduction
Chapter 6 for BST 695: Special Topics i Statistical Theory. Kui Zhag, 0 Chapter 6 Priciples of Data Reductio Sectio 6. Itroductio Goal: To summarize or reduce the data X, X,, X to get iformatio about a
More informationIntroduction to Optimization Techniques
Itroductio to Optimizatio Techiques Basic Cocepts of Aalysis - Real Aalysis, Fuctioal Aalysis 1 Basic Cocepts of Aalysis Liear Vector Spaces Defiitio: A vector space X is a set of elemets called vectors
More informationFACULTY OF MATHEMATICAL STUDIES MATHEMATICS FOR PART I ENGINEERING. Lectures
FACULTY OF MATHEMATICAL STUDIES MATHEMATICS FOR PART I ENGINEERING Lectures MODULE 5 STATISTICS II. Mea ad stadard error of sample data. Biomial distributio. Normal distributio 4. Samplig 5. Cofidece itervals
More informationExpectation and Variance of a random variable
Chapter 11 Expectatio ad Variace of a radom variable The aim of this lecture is to defie ad itroduce mathematical Expectatio ad variace of a fuctio of discrete & cotiuous radom variables ad the distributio
More information6. Kalman filter implementation for linear algebraic equations. Karhunen-Loeve decomposition
6. Kalma filter implemetatio for liear algebraic equatios. Karhue-Loeve decompositio 6.1. Solvable liear algebraic systems. Probabilistic iterpretatio. Let A be a quadratic matrix (ot obligatory osigular.
More informationProbability, Expectation Value and Uncertainty
Chapter 1 Probability, Expectatio Value ad Ucertaity We have see that the physically observable properties of a quatum system are represeted by Hermitea operators (also referred to as observables ) such
More informationON THE CAPACITY OF THE MIMO CHANNEL - A TUTORIAL INTRODUCTION - Bengt Holter
ON HE CAPACIY OF HE MIMO CHANNEL - A UORIAL INRODUCION - Begt Holter Norwegia Uiversity of Sciece ad echology Departmet of elecommuicatios O.S.Bragstads plass B, N-7491 rodheim, Norway bholter@tele.tu.o
More informationOutput Analysis and Run-Length Control
IEOR E4703: Mote Carlo Simulatio Columbia Uiversity c 2017 by Marti Haugh Output Aalysis ad Ru-Legth Cotrol I these otes we describe how the Cetral Limit Theorem ca be used to costruct approximate (1 α%
More informationDeterminants of order 2 and 3 were defined in Chapter 2 by the formulae (5.1)
5. Determiats 5.. Itroductio 5.2. Motivatio for the Choice of Axioms for a Determiat Fuctios 5.3. A Set of Axioms for a Determiat Fuctio 5.4. The Determiat of a Diagoal Matrix 5.5. The Determiat of a Upper
More informationBecause it tests for differences between multiple pairs of means in one test, it is called an omnibus test.
Math 308 Sprig 018 Classes 19 ad 0: Aalysis of Variace (ANOVA) Page 1 of 6 Itroductio ANOVA is a statistical procedure for determiig whether three or more sample meas were draw from populatios with equal
More information1 Approximating Integrals using Taylor Polynomials
Seughee Ye Ma 8: Week 7 Nov Week 7 Summary This week, we will lear how we ca approximate itegrals usig Taylor series ad umerical methods. Topics Page Approximatig Itegrals usig Taylor Polyomials. Defiitios................................................
More informationChapter 7: The z-transform. Chih-Wei Liu
Chapter 7: The -Trasform Chih-Wei Liu Outlie Itroductio The -Trasform Properties of the Regio of Covergece Properties of the -Trasform Iversio of the -Trasform The Trasfer Fuctio Causality ad Stability
More informationSieve Estimators: Consistency and Rates of Convergence
EECS 598: Statistical Learig Theory, Witer 2014 Topic 6 Sieve Estimators: Cosistecy ad Rates of Covergece Lecturer: Clayto Scott Scribe: Julia Katz-Samuels, Brado Oselio, Pi-Yu Che Disclaimer: These otes
More informationROLL CUTTING PROBLEMS UNDER STOCHASTIC DEMAND
Pacific-Asia Joural of Mathematics, Volume 5, No., Jauary-Jue 20 ROLL CUTTING PROBLEMS UNDER STOCHASTIC DEMAND SHAKEEL JAVAID, Z. H. BAKHSHI & M. M. KHALID ABSTRACT: I this paper, the roll cuttig problem
More informationL = n i, i=1. dp p n 1
Exchageable sequeces ad probabilities for probabilities 1996; modified 98 5 21 to add material o mutual iformatio; modified 98 7 21 to add Heath-Sudderth proof of de Fietti represetatio; modified 99 11
More information1.010 Uncertainty in Engineering Fall 2008
MIT OpeCourseWare http://ocw.mit.edu.00 Ucertaity i Egieerig Fall 2008 For iformatio about citig these materials or our Terms of Use, visit: http://ocw.mit.edu.terms. .00 - Brief Notes # 9 Poit ad Iterval
More informationOn the Capacity of Symmetric Gaussian Interference Channels with Feedback
O the Capacity of Symmetric Gaussia Iterferece Chaels with Feedback La V Truog Iformatio Techology Specializatio Departmet ITS FPT Uiversity, Haoi, Vietam E-mail: latv@fpteduv Hirosuke Yamamoto Dept of
More informationComplex Algorithms for Lattice Adaptive IIR Notch Filter
4th Iteratioal Coferece o Sigal Processig Systems (ICSPS ) IPCSIT vol. 58 () () IACSIT Press, Sigapore DOI:.7763/IPCSIT..V58. Complex Algorithms for Lattice Adaptive IIR Notch Filter Hog Liag +, Nig Jia
More information62. Power series Definition 16. (Power series) Given a sequence {c n }, the series. c n x n = c 0 + c 1 x + c 2 x 2 + c 3 x 3 +
62. Power series Defiitio 16. (Power series) Give a sequece {c }, the series c x = c 0 + c 1 x + c 2 x 2 + c 3 x 3 + is called a power series i the variable x. The umbers c are called the coefficiets of
More informationECON 3150/4150, Spring term Lecture 3
Itroductio Fidig the best fit by regressio Residuals ad R-sq Regressio ad causality Summary ad ext step ECON 3150/4150, Sprig term 2014. Lecture 3 Ragar Nymoe Uiversity of Oslo 21 Jauary 2014 1 / 30 Itroductio
More information