Channel Estimation of MIMO Relay Systems With Multiple Relay Nodes

Size: px
Start display at page:

Download "Channel Estimation of MIMO Relay Systems With Multiple Relay Nodes"

Transcription

1 Receved September 21, 2017, accepted November 12, 2017, date of publcaton November 20, 2017, date of current verson December 22, Dgtal Object Identfer /ACCESS Channel Estmaton of MIMO Relay Systems Wth Multple Relay Nodes ZHIQIANG HE 1,2, (Member, IEEE, JIAOLONG YANG 1, XIAODAN WANG 1, YANG LIU 1, AND YUE RONG 3, (Senor Member, IEEE 1 Key Laboratory of Unversal Wreless Communcaton, Mnstry of Educaton, Bejng Unversty of Posts and Telecommuncatons, Bejng , Chna 2 Key Laboratory of Underwater Acoustc Communcaton and Marne Informaton Technology, Mnstry of Educaton, Xamen Unversty, Xamen , Chna 3 Department of Electrcal and Computer Engneerng, Curtn Unversty, Bentley, WA 6102, Australa Correspondng author: Yue Rong (y.rong@curtn.edu.au Ths work was supported n part by the Natonal Natural Scence Foundaton of Chna under Grant and n part by the Australan Research Councl s Dscovery Projects fundng scheme under Grant DP ABSTRACT In ths paper, we consder the channel estmaton problem for multple-nput multple-output wreless relay communcaton systems wth multple relay nodes. In partcular, all ndvdual channel matrces of the frst-hop and second-hop lnks are estmated at the destnaton node by applyng the supermposed channel tranng algorthm, where tranng sequences are supermposed at the relay nodes to assst the estmaton of the relays-destnaton channel matrces. To mprove the performance of channel estmaton, we consder the estmaton error nherted from the second-hop channel estmaton and develop a new mnmal mean-squared error-based algorthm to estmate the frst-hop channel matrces. Furthermore, we derve the optmal power allocaton and tranng sequences at the source and relay nodes. Numercal examples demonstrate a better performance of the proposed supermposed channel tranng algorthm. INDEX TERMS Channel estmaton, MIMO relay, MMSE, multple relay nodes, power allocaton, supermposed tranng. I. INTRODUCTION Multple-nput multple-output (MIMO relay communcaton systems have attracted much research nterest recently due to ther capablty n mprovng the relablty and coverage of wreless systems 1, 2. Many studes have been carred out on the transcever optmzaton of amplfyand-forward (AF MIMO relay systems. The optmal relay precodng matrx whch maxmzes the source-destnaton mutual nformaton (MI of a two-hop MIMO relay system was developed n 3 and 4. In 5 and 6, the relay precodng matrx was desgned to mnmze the mean-squared error (MSE of the sgnal waveform estmaton. A unfed framework was developed n 7 and 8 to jontly optmze the source and relay precodng matrces for a broad class of objectve functons. The works n 3 8 focused on MIMO relay systems wth a sngle relay node at each hop. Systems wth multple parallel relay nodes have also attracted great nterests 9, 10. Behbahan et al. 9 developed the optmal relay precodng matrces wth multple relay nodes. In 10, jont source and relay precodng matrces optmzaton has been nvestgated wth power constrant at the output of the second-hop channel consderng both lnear and nonlnear recevers. It can be seen that for MIMO relay systems dscussed n 3 10, the knowledge of the nstantaneous channel state nformaton (CSI s requred for desgnng the optmal lnear recever at the destnaton node and optmzng the system through precodng matrces desgn. Nevertheless, n real relay communcaton systems, the nstantaneous CSI s unknown at the destnaton node, and therefore, needs to be estmated. Lolou and Vberg 11 and Lolou et al. 12 developed a least-squares (LS based channel estmaton algorthm for MIMO relay systems and further mproved ts performance by the weghted LS (WLS approach. A two-stage channel tranng method was developed n 13, where the optmal tranng sequence at the source and relay nodes was derved. In 14, a parallel factor (PARAFAC analyss based MIMO relay channel estmaton algorthm was proposed. Supermposed channel tranng algorthms were developed n 15 and 16 for two-way MIMO relay systems, and n 17 for tme-varyng MIMO relay systems, where a tranng sequence s supermposed at the relay node to estmate VOLUME 5, IEEE. Translatons and content mnng are permtted for academc research only. Personal use s also permtted, but republcaton/redstrbuton requres IEEE permsson. See for more nformaton

2 the CSI of frst-hop and second-hop channel at the destnaton nodes. The channel estmaton algorthms n were developed for MIMO relay systems wth a sngle relay node at each hop 3 8. Obvously, channel estmaton becomes more challengng for systems wth multple parallel relay nodes as more unknowns need to be estmated. In 18, LS and maxmum lkelhood (ML based channel estmators were derved based on the parameterzaton of the estmaton problem. The ML estmaton technque was used to estmate the bass expanson model (BEM weghtng coeffcents n 19. However, the optmzaton of tranng sequence, whch may further mprove the performance of the estmators, was not dscussed n 18 and 19. In ths paper, we apply the prncple of supermposed channel tranng to estmate all ndvdual frst-hop and secondhop channel matrces of MIMO relay systems wth multple parallel relay nodes. Note that the knowledge of both the source-relay and relay-destnaton channels s requred at the destnaton node for developng the optmal recever 9, 10. In the proposed algorthm, the channel estmaton process s completed n two tme blocks. At the frst tme block, the source node broadcasts tranng sequence to all relay nodes, whle at the second tme block, each relay node transmts lnearly amplfed receved sgnals as well as ts own tranng sequence. Then all ndvdual channel matrces of the frsthop and second-hop lnks can be effcently estmated at the destnaton node. We derve the optmal structure of the tranng sequences that mnmze the sum MSE of channel estmaton and optmze the power allocaton between the tranng sequences at the source and relay nodes. We consder the estmaton error nherted from the estmaton of the second-hop channel matrces and develop a new mnmal MSE (MMSE- based algorthm to estmate the frst-hop channel matrces, for whch no effcent method was proposed n 15. Moreover, compared wth 15 and 16, the optmzaton problems n ths paper are more challengng to solve, as more channel tranng sequences need to be optmzed. Numercal examples demonstrate a better performance of the proposed supermposed channel tranng algorthm compared wth the conventonal two-stage channel estmaton approach for MIMO relay systems wth multple relay nodes. The rest of ths paper s organzed as follows. The system model of a two-hop MIMO relay communcaton system wth multple relay nodes s presented n Secton II. The supermposed channel tranng algorthm s developed n Secton III, where the optmal tranng sequences and power allocaton at the source and relay nodes are derved. In Secton IV, we present a new MMSE-based algorthm to estmate the frsthop channel matrces consderng the error of the second-hop channel estmaton. Secton V shows numercal smulaton results to demonstrate the performance of the proposed algorthms. Fnally, conclusons are drawn n Secton VI. The followng notatons are used throughout the paper. Vectors and matrces are represented by lower case and upper case bold letters, respectvely; Superscrpts ( T, ( H, and ( 1 denote transpose, Hermtan transpose, and matrx nverson, respectvely; denotes the matrx Kronecker product 20; tr( stands for the matrx trace; vec( denotes the vectorzaton operator whch stacks all column vectors of a matrx on top of each other; I n denotes an n n dentty matrx; 1 s a vector wth all one entres; bdag and dag( stand for a block dagonal and dagonal matrx, respectvely; and E denotes the statstcal expectaton. II. SYSTEM MODEL We consder a two-hop MIMO relay communcaton system whch conssts of one source node, K parallel relay nodes, and one destnaton node as llustrated n Fg. 1. The source and the destnaton nodes have N s and N d antennas, respectvely, and the th relay node has N antennas. For = 1,..., K, H denotes the N N s frst-hop channel matrx from the source node to the th relay node and G s the N d N secondhop channel matrx from the th relay node to the destnaton node. Due to ts mert of smplcty, we consder the AF relay scheme at each relay. FIGURE 1. Block dagram of a two-hop MIMO relay communcaton system wth multple parallel relay nodes. The channel tranng process s mplemented n two tme blocks. At the frst tme block, the source node sends an N s T channel tranng matrx S to all relay nodes, where T s the length of the tranng sequence. The sgnals receved at the th relay node can be wrtten as Y = H S + V, = 1,..., K (1 where Y and V are the N T receved sgnal matrx and nose matrx at the th relay node, respectvely. At the second tme block, the source node s slent, and the th relay node amplfes ts receved sgnal wth an amplfyng factor α > 0 and supermposes ts own tranng matrx T. Therefore, the sgnal matrx transmtted by the th relay node s gven by X = αy + T, = 1,..., K. (2 The receved sgnal at the destnaton node can be wrtten as Y d = K G X + V d ( VOLUME 5, 2017

3 where V d s the N d T addtve Gaussan nose matrx at the destnaton node. We assume that the channel matrces H and G satsfy the well-known Gaussan-Kronecker model 21, where H and G are complex Gaussan random matrces wth H CN (0, T s R, G CN (0, C R d (4 where T s and R denote the N s N s and N N covarance matrx at the transmt and receve sde of H, respectvely, whle C and R d denote the N N and N d N d covarance matrx at the transmt and receve sde of G, respectvely. In other words, from (4 we have H = A H,w B H s, G = A d G,w K H (5 where A A H = R, B s B H s = T T s, A da H d = R d, K K H = C T, = 1,..., K, H,w and G,w are N N s and N d N complex Gaussan random matrces wth ndependent and dentcally dstrbuted (..d. zero mean and unt varance entres. We assume that H,w and G,w, = 1,..., K, are statstcally ndependent of each other. We also assume that the knowledge on the channel covarance matrces s avalable at the destnaton node. By substtutng (1 and (2 nto (3, we can rewrte the receved sgnal at the destnaton node as K Y d = G α(h S + V + T + V d = αms + K G T + V (6 where V = K αg V +V d s the total nose matrx at the destnaton node and M = K G H can be vewed as the compound source-destnaton channel matrx. It can be seen from (6 that the frst-hop channel H s multpled wth the second-hop channel G, whch makes t dffcult to estmate both H and G n one step. Thus, we propose an effcent twostep channel estmaton algorthm to estmate the second-hop channel G frst based on T, and then estmate the frst-hop channels. III. SUPERIMPOSED CHANNEL TRAINING FOR THE SECOND-HOP CHANNEL ESTIMATION In ths secton, the supermposed channel tranng prncple s appled to estmate the channel matrces M and G, = 1,..., K. Moreover, we desgn the optmal tranng matrces S and T, = 1,..., K, and the relay amplfyng factor α to mnmze the sum MSE of the channel estmaton. The man dea of the supermposed channel tranng algorthm s to use T to estmate the second-hop channel matrces G, = 1,..., K. By ntroducng the egenvalue decompostons (EVDs of T T s = U s s U H s, CT = U U H, = 1,..., K we have B H s = s 1 2 s U H s, KH = 1 2 U H, = 1,..., K (7 where s and are arbtrary N s N s and N N untary matrx, respectvely. Based on (5, the receved sgnal at the destnaton node (6 can be rewrtten as where M Y d = α M S + K G H K G T + V (8 S U H s S, H H U s, = 1,..., K G G U, T U H T, = 1,..., K. (9 In the followng, we develop an algorthm to estmate M and G n (8. In Secton IV, we wll present the algorthm to estmate the frst-hop channels H from (8. It wll be shown n Secton IV that the MSE of estmatng H ncreases wth the MSE of estmatng M and G. Thus, t s mportant to optmze the tranng matrces S and T, = 1,..., K, to mnmze the MSE of estmatng H. Applyng the dentty of vec(abc = (C T Avec(B 20, we can rewrte (8 n vector form as α y d = S T I Nd, T T 1 I N d,..., T T K I N d T m T, g T 1,..., gt K + v = Lγ + v (10 where α L S T I Nd, T T 1 I N d,..., T T K I N d T γ m T, g T 1,..., gt K y d vec(y d, m vec( M v vec(v, g vec( G, = 1,..., K. In (10, γ s the vector of unknown wth a dmenson of Q N d (N s + K N, L has a dmenson of T N d Q, and there s T N s + K N. Thanks to ts computatonal smplcty, a lnear estmator s appled at the destnaton node to estmate γ as ˆγ = W H y d (11 where ˆγ denotes the estmaton of γ and W s the weght matrx of the lnear estmator. Usng (11, the MSE of channel estmaton s gven by MSE 1 = Etr(( ˆγ γ ( ˆγ γ H = tr((w H L I Q R γ (W H L I Q H + W H R v W (12 where R γ = Eγ γ H and R v = Evv H are the covarance matrx of γ and v, respectvely. VOLUME 5,

4 Lemma 1: R γ and R v are gven by ( K R v = I T α tr(c T R d + I Nd (13 R γ = bdag s br d, 1 R d,..., K R d (14 where b K tr(r C T. Proof: See Appendx A. It s well-known that (12 s mnmzed by the lnear MMSE estmator 22 gven by ( 1LRγ W = LR γ L H + R v. (15 By substtutng (15 back nto (12 and applyng the matrx nverson lemma of (A + BCD 1 = A 1 A 1 B(DA 1 B + C 1 1 DA 1 we can obtan the MSE of estmatng γ as MSE 1 = tr ( (R 1 γ + L H R 1 v L 1. (16 The transmsson power consumed by the source node s gven by tr(ss H = tr( S S H. (17 Usng (2 and Lemma 3 n Appendx A, the transmsson power of the th relay node can be calculated as αetr(h SS H H H + I N + tr(t T H = αn + α tr( s S S H tr(r + tr( T T H. (18 From (16-(18, the optmal amplfyng factor α and the optmal tranng matrces can be desgned through solvng the followng optmzaton problem mn α, S, T 1,..., T K tr ( (R 1 γ + L H R 1 v L 1 (19 s.t. tr( S S H p s, α > 0 (20 αn + α tr( s S S H tr(r + tr( T T H p, = 1,..., K (21 where p s s the transmsson power avalable at the source node and p s the transmsson power at the th relay node. Usng 15, Th. 1, the optmal S and T as the soluton to the problem (19 (21 satsfy the equatons below for, j = 1,..., K S S H = s, T T H = (22 S T H = 0, T T H j = 0, = j (23 where s and are N s N s and N N dagonal matrx, respectvely, wth non-negatve dagonal elements. Based on (22 and (23, the optmal structure of the tranng sequences s gven by S = U s 1 2 s s, T = U 1 2, = 1,..., K (24 where, = s, 1,..., K, s an N T sem-untary matrx satsfyng the followng equatons H = I N, H j = 0,, j = s, 1,..., K, = j. (25 We would lke to note that sem-untary satsfyng (25 can be easly obtaned, for nstance, from the normalzed dscrete Fourer transform (DFT matrces. Based on (22 and (23, t can be seen that the MSE matrx (R 1 γ + L H R 1 v L 1 n (16 becomes block dagonal under the optmal S and T, ndcatng that the estmaton errors of m T, g T, = 1,..., K, are uncorrelated. By ntroducng the EVD of R d = U d d U H d, the problem (19-(21 can be rewrtten as ( K mn tr (D ds + α s D v 1 + (D d + D v 1 α,{ } (26 s.t. tr( s p s (27 α > 0, 0, = s, 1,..., K (28 αn + αtr( s s tr(r (29 + tr( p, = 1,..., K (30 where { } = {, = s, 1,..., K} and D ds = 1 s (b d 1 (31 D d = 1 1 d, = 1,..., K (32 ( K 1. D v = αtr(c T d + I Nd (33 Usng (31-(33, the problem (26-(29 can be equvalently converted to the followng optmzaton problem wth scalar varables mn α,{σ,m } s.t. N s N s N d ( n=1 N + 1 bλ s,m λ d,n + ασ s,m d v,n K N d ( n=1 1 1 λ,m λ d,n + σ,m d v,n 1 (34 σ s,m p s (35 ( Ns α λ s,m σ s,m tr(r + N N + σ,m p, = 1,..., K (36 α > 0, σ,m 0, m = 1,..., N, = s, 1,..., K (37 where {σ,m } = {σ,m, m = 1,..., N, = s, 1,..., K}, λ d,n s the nth dagonal element of d, λ,m, σ,m, = s, 1,..., K, are the mth dagonal element of and, respectvely, and ( K 1 d v,n αtr(c T λ d,n + 1. (38 For a gven α, the problem (34 (37 s a convex optmzaton problem wth respect to {σ,m }. Ths s because VOLUME 5, 2017

5 when α s fxed, b, λ s,m, λ d,n, d v,n and λ,m are known varables, and (34 s monotoncally decreasng and convex wth respect to {σ,m }. Moreover, the constrants n (35 and (36 are lnear nequalty constrants when α s fxed. Thus, wth fxed α, the problem (34 (37 wth respect to {σ,m } s a convex optmzaton problem, whch can be effcently solved by the Lagrange multpler method. The Karush-Kuhn-Tucker (KKT optmalty condtons 23 of the problem (34 (37 are gven below. Frstly, the gradent condtons are gven by N d n=1 N d αd v,n K ((bλ s,m λ d,n 1 + ασ s,m d v,n 2 = µ s + µ e,m (39 m = 1,..., N s d v,n n=1 ((λ,m λ d,n 1 + σ,m d v,n 2 = µ m = 1,..., N, = 1,..., K (40 where e,m = αλ s,m tr(r, and µ 0, = s, 1,..., K, are the Lagrange multplers. Secondly, the complementary slackness condtons are gven by N s µ s (p s σ s,m = 0 µ (p α ( Ns = 1,..., K. λ s,m σ s,m tr(r + N σ,m = 0 N When α and µ, = s, 1,..., K, are fxed, the nonnegatve {σ s,m } can be obtaned through the b-secton search, as the left-hand sde (LHS of (39 and (40 are monotoncally decreasng functons of σ s,m and σ,m, respectvely. To obtan µ s and µ, we can apply an outer b-secton search loop snce the LHS of (35 s a decreasng functon of σ s,m, and the LHS of (36 s an ncreasng functon of σ s,m and σ,m. Meanwhle, n (39, σ s,m monotoncally decreases wth respect to µ s and µ, and σ,m s a monotoncally decreasng functon of µ n (40. When α also becomes a varable n the optmzaton, the problem (34-(37 s not a convex optmzaton problem. But n ths case, the followng lemma s n order. Lemma 2: The object functon (34 subjectng to constrants (35-(37 s a unmodal functon wth respect to α. Proof: See Appendx B. For a unmodal functon, ts mnmal value can be effcently obtaned by usng the golden secton search (GSS 24. The procedure of applyng the GSS technque to fnd the optmal α s descrbed n Table I, where φ > 0 s the reducton factor, denotes the absolute value, and ε s a postve constant close to 0. It s shown n 24 that the optmal φ s 1.618, also known as the golden rato. It can be observed from Table I that at each teraton, the GSS method reduces the nterval contanng the optmal α to tmes of nterval at the precedng teraton. Thus, TABLE 1. Algorthm I: Procedure of fndng the optmal α n the Problem (34 (37. the length of the nterval contanng the optmal α after the nth teraton s Ɣ n = n Ɣ 0 24, where Ɣ 0 = α u α l s the length of the ntal feasble nterval. Therefore, the number of teratons requred to acheve the desred accuracy ε n Table 1 s gven by N = ln ε ln Ɣ 0 ln 0.618, where denotes the celng operator. The value of ε can be determned by consderng the complexty-performance tradeoff. In the smulatons, we choose ε = 10 3 and a l, a u = 0, 1. Thus, Ɣ 0 = 1 and N = 15. The unmodalty of (34 wth respect to α subjectng to (35-(37 wthn bounds a l, a u = 0, 1 s llustrated n Fg. 7 n Appendx B. IV. MMSE-BASED FIRST-HOP CHANNEL ESTIMATION In ths secton, we propose an MMSE-based algorthm to estmate the frst-hop channel matrces consderng the error of the second-hop channel estmaton n Secton III. After estmatng the second-hop channels G, = 1,..., K, ˆ G T can be subtracted from (8 as Ȳ d = Y d K ˆ G T = αg H S + K ( G ˆ G T + V = αĝ H S + V (41 where H = H T 1,..., H T K T and V = α(g Ĝ H S + K ( G ˆ G T + V (42 s the total nose matrx ncludng the channel estmaton errors. By applyng the vectorzaton operaton to both sdes of (41, we have ȳ d = ( α S T Ĝ h + v = L 2 h + v (43 where ȳ d = vec(ȳ d, h = vec( H, v = vec( V, and L 2 = α S T Ĝ. From (43, an MMSE estmate of h can be obtaned by ˆ h = W H 2 ȳd (44 where W 2 = (L 2 R h LH 2 + R v 1 L 2 R h, R h = E h h H, and R v = E v v H. From (44, the MSE of estmatng h s gven VOLUME 5,

6 FIGURE 2. Example 1: NMSE versus p at varous α wth N d = N = 4 and ρ = 0.8. by ( (R 1 MSE 3 = tr h + LH 2 R 1 v L 1 2 (45 where the dervaton of R h and R v s shown n Appendx C. Note that n (45, we take nto account the equvalent estmaton error nherted from the second-hop channel estmaton,.e., R v. Therefore, the total MSE of the frst-hop channel matrces estmaton depends on the accuracy of the second-hop channel matrces estmaton. In other words, the total MSE n (45 ncreases wth the ncreasng of the MSE of the second-hop channel estmaton. V. NUMERICAL EXAMPLES In ths secton, we study the performance of the proposed supermposed channel tranng algorthm for MIMO relay systems wth multple relay nodes through numercal smulatons. We smulate a MIMO relay system wth K = 2 relay nodes whch have the same number of antennas as the source node,.e., N s = N = N, = 1, 2, and we choose the smallest T possble,.e., T = N s + K N = 3N. The optmal tranng matrces for the supermposed channel tranng method are generated accordng to (24. In partcular, the sem-untary matrces n (25 are generated based on the normalzed DFT matrces as s m,n = 1 2πmn j 3N e 3N, 1 m,n = 1 2π(m+Nn j e 3N, 3N 2 m,n = 1 2π(m+2Nn j e 3N, 3N m = 1,..., N, n = 1,..., 3N. The channel covarance matrces have the wdely used exponental Toepltz structure 21 such that T s = R = C = ρ m n, = 1, 2, m, n = 1,..., N, and R d = ρ m n, m, n = 1,..., N d, where ρ s the correlaton coeffcent wth magntude ρ < 1. Wthout loss of generalty, we consder ρ = 0.8 and ρ = 0.2 for the hgh and low channel correlaton scenaros, respectvely. We assume that all nodes have the same amount of transmsson power as p s = p 1 = p 2 = p. In all smulaton examples, the normalzed MSE (NMSE of channel estmaton at the destnaton node s computed. In the frst example, we compare the NMSE performance of the supermposed channel tranng algorthm at varous α. Fg. 2 shows the NMSE of the proposed algorthm versus p at varous α wth N d = N = 4 and ρ = 0.8. The curve of the optmal α s obtaned by applyng the GSS method n the proposed supermposed channel tranng algorthm to fnd the optmal α for each value of p. It can be observed from Fg. 2 that the GSS technque s an effcent method to fnd the optmal α snce the optmal α curve constantly has the lowest MSE level for all values of p. Furthermore, we observe from Fg. 2 that the optmal α vares wth respect to p, whch means that usng a fxed α s strctly suboptmal. In partcular, for p between 10dB and 30dB, the NMSE wth α = 0.2 s close to the NMSE usng the optmal α. However, the curve assocated wth α = 0.07 has a lower NMSE than the one assocated wth α = 0.2 when p = 5dB. In addton, for other smulaton examples (e.g. dfferent N and ρ, the NMSE wth α = 0.2 mght not be close to the NMSE usng the optmal α. FIGURE 3. Example 2: NMSE versus p at varous N wth ρ = 0.8 and N d = N. In the second example, we set N d = N and nvestgate the NMSE performance of the proposed supermposed channel tranng algorthm usng the optmal α for varous smulaton scenaros. A comparson of the proposed algorthm has been made wth the conventonal two-stage MMSE-based channel tranng algorthm, where random orthogonal plot sequences are adopted to estmate channel matrces wth equally dstrbuted transmsson power at the relay node between two stages 15. Fg. 3 shows the NMSE performance of both methods wth ρ = 0.8 for N = 2 and N = 4, whle Fg. 4 demonstrates the performance of two methods wth ρ = 0.2. It can be observed from Fgs. 3 and 4 that the two algorthms have smlar performance when p < 10dB, whle at a hgh power level, the supermposed channel tranng algorthm yelds much smaller estmaton error than the conventonal two-stage channel estmaton method. Moreover, the relatonshp between the curves almost stays the same when the correlaton coeffcent ρ changes, ndcatng that the VOLUME 5, 2017

7 FIGURE 4. Example 2: NMSE versus p at varous N wth ρ = 0.2 and N d = N. proposed algorthm s effcent n both scenaros of hgh and low channel correlaton. We also observe from Fgs. 3 and 4 that for both algorthms the NMSE ncreases wth N, as more unknowns need to be estmated wth a larger number of antennas. In the thrd example, we study the NMSE performance of the algorthm proposed n Secton IV whch retreves the ndvdual CSI of the frst-hop channels. Snce the frst-hop channel estmaton s based on the second-hop one, and the error n second-hop estmaton s propagated to the frsthop channel estmaton, we apply the proposed supermposed channel tranng algorthm n Secton III to estmate the second-hop channels n ths example, whch has a better NMSE performance than the conventonal two-stage algorthm. Fg. 5 shows the NMSE performance of ths method wth ρ = 0.2 for varous N and N d. It can be seen that wth fxed N d, the NMSE ncreases wth N, as more unknowns need to be estmated. As expected, for a gven N, the NMSE decreases wth N d snce more observatons are avalable. FIGURE 5. Example 3: NMSE of the frst-hop channel estmaton versus p at varous N d and N wth ρ = 0.2. FIGURE 6. Example 3: NMSE of the frst-hop channel estmaton versus p at varous N d and N wth ρ = 0.8. Fg. 6 demonstrates the performance of ths approach wth ρ = 0.8. Smlar to Fg. 5, t can be seen from Fg. 6 that a better MSE performance can be acheved by settng N d = N 1 + N 2 nstead of N d = N 1 = N 2 = N. VI. CONCLUSIONS In ths paper, we have appled the supermposed channel tranng approach to MIMO relay communcaton systems wth multple relay nodes. The channel matrces of both the frst-hop and the second-hop lnks can be effcently estmated usng the proposed algorthms. The optmal tranng sequences and power allocaton at the source and relay nodes are derved. Furthermore, a new MMSE-based estmator has been developed to retreve the frst-hop channel nformaton. A better performance of the proposed supermposed channel tranng algorthm has been shown through numercal examples. The channel tranng algorthm proposed n ths paper can be readly extended to two-way MIMO relay networks wth multple relay nodes. APPENDIX A PROOF OF LEMMA 1 In order to derve R v and R γ, here we ntroduce the followng lemma. Lemma 3 25: For H CN (0,, there s EHAH H = tr(a T and EH H AH = tr( A T. To derve R v, we frst calculate the covarance matrx of v m usng (4 and Lemma 3, where v m denotes the mth column of V K E v m v H m = E αg v,m v H,m GH + v d,m v H d,m K = E αg G H + I Nd = K αtr(c T R d + I Nd (46 VOLUME 5,

8 where v,m and v d,m are the mth columns of V and V d, respectvely. As v m, m = 1,..., T, are ndependent, R v can be wrtten as ( K R v = I T αtr(c T R d + I Nd. Now let us calculate R γ. From (5, (7, and (9, the mth column of M and G can be wrtten as m m = λ 1 2 s,m K G A H,w π s,m, g,m = λ 1 2,m A d G,w π,m where π s,m and π,m are the mth column of s and, respectvely, λ s,m and λ,m are the mth dagonal element of s and, respectvely. Then smlar to (46, the covarance matrx of m m and g,m are gven by K E m m m H m = λ s,me G A H,w π s,m π H s,m HH,w AH G H K = λ s,m tr(r C T R d = λ s,m br d E g,m g H,m = λ,me A d G,w π,m π H,m GH,w AH d = λ,m R d. As m m, m = 1,..., N s and g,m, m = 1,..., N, = 1,..., K, are ndependent, we have R γ = bdag s br d, 1 R d,..., K R d. APPENDIX B PROOF OF LEMMA 2 Denotng a m,n 1/ ( bλ s,m λ d,n, cm ασ s,m, and g,m,n 1/ ( λ,m λ d,n, we can rewrte the problem (34 (37 as mn α,c,{σ } N s N d n=1 N + 1 a m,n + c m d v,n K N d n=1 1 g,m,n + σ,m d v,n (47 s.t. 1 T c αp s (48 z T c + 1T σ p αn, = 1,..., K (49 α > 0, c m 0, m = 1,..., N s (50 σ,m 0, m = 1,..., N, = 1,..., K (51 where c c 1,..., c Ns T, z tr(r λ s,1,..., λ s,ns T, σ σ,1,..., σ,n T, and {σ } = {σ, = 1,..., K}. For a very small α, the value of (47 s manly governed by the constrant n (48, snce the constrants n (49 s not actve compared wth the constrant (48 when α s small. As α ncreases from a small value, the feasble regon determned by (48 expands, whch contrbutes to the decrease n the value of (47. On the other hand, when α s large, the value of (47 s manly affected by the constrants n (49, as the constrant (48 s not actve compared wth those n (49 for a large value of α. As α decreases from a large value, the feasble regon specfed by (49 expands, resultng n a decreasng of (47. Now let us consder the effect of α on d v,n. Note that d v,n n (38 s monotoncally decreasng when α ncreases, and the value of (47 ncreases wth decreasng d v,n. Takng the two effects above nto account, we can draw the followng concluson regardng the value of (47 wth respect to α. When α ncreases from a very small postve number, the value of (47 starts to decrease as the potental decrease of (47 caused by the relaxed feasble regon (48 domnates the potental ncrease of (47 due to the decreasng d v,n. The value of (47 keeps decreasng as α ncreases tll a turnng pont where the decreasng of d v,n starts to domnate the effect of relaxed feasble regon (48. After ths turnng pont, the value of (47 wll monotoncally ncrease wth an ncreasng α. Therefore, the objectve functon (34 subjectng to (35-(37 s proved to be a unmodal functon wth respect to α. To valdate the analyss above, we plot the MSE value (34 versus α n Fg. 7 at varous levels of p s. Here we set p 1 = p 2 = 20dB and the other smulaton setups are the same as those n Fg. 2. It can be clearly seen from Fg. 7 that the objectve functon (34 subjectng to the constrants (35 (37 s a unmodal functon of α. FIGURE 7. NMSE versus α at varous p s wth p 1 = p 2 = 20dB. APPENDIX C CALCULATION OF R h AND R v From (5 and (7, we have H = H U s = A H,w s 1 2 s. (52 Usng (52, the covarance matrx of the pth column of H, denoted as h,p, s gven by E h,p h H,p = λ s,pe A H,w π s,p π H s,p HH,w AH = λ s,p R. (53 Then the covarance matrx of h p = h T 1,p,..., h T K,p T, the pth column of H, can be obtaned based on (53 as R h p = bdag λ s,p R 1, λ s,p R 2,..., λ s,p R K. ( VOLUME 5, 2017

9 As h = h T 1,, h T N s T, we have R h = E h h H = bdag R h 1, R h 2,..., R h Ns. Now we start to calculate R v = E v v H. From (42, the correlaton matrx of the mth and nth columns of V, m, n = 1,..., T, s gven by E v m v H n = αe H s m s H n H H H K + E t,m t H,n H + Ev m v H n (55 where = ( 1,..., K, = U H, = G ˆ G, = 1,..., K, ˆ G s the estmate of G, s m, t,m, and v m are the mth columns of S, T, and V, respectvely. From (46, we obtan K Ev m v H n = αtr(c T R d + I Nd, m = n; 0, m = n. 2 Wth t,m = 1 φ,m, where φ,m s the mth column of, the (e, f -th entry of the second term n (55 can be calculated as K E t,m t H,n H = = K K e,f E (δ,1,e,..., δ,n,e 1 2 φ,m φ H,n 1 ( 2 H δ,1,f,..., δ,n,f tr( 1 2 φ,m φ H,n 1 2 D,e,f (56 where δ,p,q s the qth entry of δ,p = g,p ˆ g,p, p = 1,..., N, q = 1,..., N d, g,p and ˆ g,p denote the pth columns of G and ˆ G, respectvely, and D,e,f s gven by (δ,1,f H ( D,e,f = E,..., δ,n,f δ,1,e,..., δ,n,e = dag(eδ,1,e δ,1,f,..., Eδ,N,eδ,N,f = dag (,1 e,f,...,,n e,f. (57 Here,p = E δ,p δ H,p = E ( g,p ˆ g,p ( g,p ˆ g H,p ( K 1 = (λ,p R d 1 + σ,p αtr(c T R d + I Nd 1. (58 In other words, D,e,f n (57 s a dagonal matrx whose pth dagonal entry, p = 1,..., N, s the (e, f -th entry of,p n (58. Fnally, the frst term n (55 can be calculated as K E H s m s H n H K H H = E H s m s H n H H H = where we used the fact that K E H s m s H n H H H = s H n K s s m E R H (59 E H s m s H n H H = EA H,w s 1 2 s s m s H n 1 2 s H s HH,w AH = s H n s s m R. To derve E R H, ts (e, f -th entry, e, f = 1,..., N d, can be calculated as E R H e,f (δ,1,e ( = E,..., δ,n,e U H H R U δ,1,f,..., δ,n,f = tr(u H R U D,e,f. (60 REFERENCES 1 Y. Fan and J. Thompson, MIMO confguratons for relay channels: Theory and practce, IEEE Trans. Wreless Commun., vol. 6, no. 5, pp , May L. Sangunett, A. A. D Amco, and Y. Rong, A tutoral on the optmzaton of amplfy-and-forward MIMO relay systems, IEEE J. Sel. Areas Commun., vol. 30, no. 8, pp , Sep X. Tang and Y. Hua, Optmal desgn of non-regeneratve MIMO wreless relays, IEEE Trans. Wreless Commun., vol. 6, no. 4, pp , Apr O. Muñoz-Medna, J. Vdal, and A. Agustín, Lnear transcever desgn n nonregeneratve relays wth channel state nformaton, IEEE Trans. Sgnal Process., vol. 55, no. 6, pp , Jun W. Guan and H. Luo, Jont MMSE transcever desgn n non-regeneratve MIMO relay systems, IEEE Commun. Lett., vol. 12, no. 7, pp , Jul Y. Rong, Lnear non-regeneratve multcarrer MIMO relay communcatons based on MMSE crteron, IEEE Trans. Commun., vol. 58, no. 7, pp , Jul Y. Rong, X. Tang, and Y. Hua, A unfed framework for optmzng lnear nonregeneratve multcarrer MIMO relay communcaton systems, IEEE Trans. Sgnal Process., vol. 57, no. 12, pp , Dec Y. Rong and Y. Hua, Optmalty of dagonalzaton of mult-hop MIMO relays, IEEE Trans. Wreless Commun., vol. 8, no. 12, pp , Dec A. S. Behbahan, R. Merched, and A. M. Eltawl, Optmzatons of a MIMO relay network, IEEE Trans. Sgnal Process., vol. 56, no. 10, pp , Oct A. Todng, M. R. A. Khandaker, and Y. Rong, Jont source and relay optmzaton for parallel MIMO relay networks, EURASIP J. Adv. Sgnal Process., vol. 2012, p. 174, Aug P. Lolou and M. Vberg, Least-squares based channel estmaton for MIMO relays, n Proc. Int. ITG Workshop Smart Antennas, Feb. 2008, pp P. Lolou, M. Vberg, and M. Coldrey, Effcent channel estmaton technques for amplfy and forward relayng systems, IEEE Trans. Commun., vol. 60, no. 11, pp , Nov T. Kong and Y. Hua, Optmal desgn of source and relay plots for MIMO relay channel estmaton, IEEE Trans. Sgnal Process., vol. 59, no. 9, pp , Sep VOLUME 5,

10 Z. He et al.: Channel Estmaton of MIMO Relay Systems Wth Multple Relay Nodes 14 Y. Rong, M. R. A. Khandaker, and Y. Xang, Channel estmaton of dualhop MIMO relay system va parallel factor analyss, IEEE Trans. Wreless Commun., vol. 11, no. 6, pp , Jun C. W. R. Chong, Y. Rong, and Y. Xang, Channel tranng algorthms for two-way MIMO relay systems, IEEE Trans. Sgnal Process., vol. 61, no. 16, pp , Aug C. W. R. Chong, Y. Rong, and Y. Xang, Channel estmaton for two-way MIMO relay systems n frequency-selectve fadng envronments, IEEE Trans. Wreless Commun., vol. 14, no. 1, pp , Jan C. W. R. Chong, Y. Rong, and Y. Xang, Channel estmaton for tmevaryng MIMO relay systems, IEEE Trans. Wreless Commun., vol. 14, no. 12, pp , Dec H. Mehrpouyan, S. D. Blosten, and B. Ottersten, Smultaneous estmaton of mult-relay MIMO channels, n Proc. IEEE Global Commun. Conf., Atlanta, GA, USA, Dec. 2013, pp L. Canonne-Velasquez, H. Nguyen-Le, and T. Le-Ngoc, Cascaded doubly-selectve channel estmaton n mult-relay AF OFDM transmssons, n Proc. IEEE Int. Conf. Commun., Ottawa, ON, Canada, Jun. 2012, pp J. W. Brewer, Kronecker products and matrx calculus n system theory, IEEE Trans. Crcuts Syst., vol. CAS-25, no. 9, pp , Sep D. S. Shu, G. Foschn, M. Gans, and J. Kahn, Fadng correlaton and ts effect on the capacty of multelement antenna systems, IEEE Trans. Commun., vol. 48, no. 3, pp , Mar S. M. Kay, Fundamentals of Statstcal Sgnal Processng: Estmaton Theory, vol. 1. Englewood Clffs, NJ, USA: Prentce-Hall, S. Boyd and L. Vandenberghe, Convex Optmzaton. Cambrdge, U.K.: Cambrdge Unv. Press, A. Antonou and W.-S. Lu, Practcal Optmzaton: Algorthms and Engneerng Applcatons. Sprng Street, NY, USA: Sprnger, A. Gupta and D. Nagar, Matrx Varate Dstrbutons. London, U.K.: Chapman & Hall, XIAODAN WANG receved the B.E. and M.Sc. degrees n nformaton and communcaton engneerng from the Bejng Unversty of Posts and Telecommuncatons, Chna, n 2014 and 2017, respectvely. She majors n new physcal layer technology appled n 5G, wth an emphass on channel estmaton and dfferent schemes of channel codng and nonorthogonal multple access. ZHIQIANG HE (S 01 M 04 receved the B.E. degree n sgnal and nformaton processng and Ph.D. degree (Hons. n sgnal and nformaton processng from the Bejng Unversty of Posts and Telecommuncatons, Chna, n 1999 and 2004, respectvely. Snce 2004, he has been wth the School of Informaton and Communcaton Engneerng, Bejng Unversty of Posts and Telecommuncatons, where he s currently a Professor and the Drector of the Center of Informaton Theory and Technology. Hs research nterests nclude sgnal and nformaton processng n wreless communcatons, networkng archtecture and protocol desgn, machne learnng, and underwater acoustc communcatons. YUE RONG (S 03 M 06 SM 11 receved the Ph.D. degree (summa cum laude n electrcal engneerng from the Darmstadt Unversty of Technology, Darmstadt, Germany, n He was a Post-Doctoral Researcher wth the Department of Electrcal Engneerng, Unversty of Calforna, Rversde, from 2006 to Snce 2007, he has been wth the Department of Electrcal and Computer Engneerng, Curtn Unversty, Bentley, WA, Australa, where he s currently a Professor. Hs research nterests nclude sgnal processng for communcatons, wreless communcatons, underwater acoustc communcatons, applcatons of lnear algebra and optmzaton methods, and statstcal and array sgnal processng. He has publshed over 150 journal and conference papers n these areas. Dr. Rong was a recpent of the Best Paper Award at the 2011 Internatonal Conference on Wreless Communcatons and Sgnal Processng, the Best Paper Award at the 2010 Asa-Pacfc Conference on Communcatons, and the Young Researcher of the Year Award of the Faculty of Scence and Engneerng at Curtn Unversty n He was an Edtor of the IEEE WIRELESS COMMUNICATIONS LETTERS from 2012 to 2014, a Guest Edtor of the IEEE JOURNAL ON SELECTED AREAS IN COMMUNICATIONS specal ssue on theores and methods for advanced wreless relays, and a TPC Member for the IEEE ICC, WCSP, IWCMC, EUSIPCO, and ChnaCom. He s an Assocate Edtor of the IEEE TRANSACTIONS ON SIGNAL PROCESSING. JIAOLONG YANG receved the B.E. degree n nformaton engneerng from the Bejng Unversty of Posts and Telecommuncatons, Bejng, Chna, n 2015, where he s currently pursung the M.Sc. degree n electroncs and telecommuncaton engneerng. Hs research nterests are manly on wreless communcatons YANG LIU receved the M.Sc. degree n nformaton and communcaton engneerng from the Bejng Unversty of Posts and Telecommuncatons, Bejng, Chna, n Hs research focuses on frequency-doman equalzaton algorthms, and the results have been publshed at IC-NIDC. VOLUME 5, 2017

Joint source and relay design for MIMO multi-relay systems using projected gradient approach

Joint source and relay design for MIMO multi-relay systems using projected gradient approach Todng et al EURASIP Journal on Wreless Communcatons and Networkng 04, 04:5 http://jwcneuraspjournalscom/content/04//5 RESEARC Open Access Jont source and relay desgn for MIMO mult-relay systems usng projected

More information

ECE559VV Project Report

ECE559VV Project Report ECE559VV Project Report (Supplementary Notes Loc Xuan Bu I. MAX SUM-RATE SCHEDULING: THE UPLINK CASE We have seen (n the presentaton that, for downlnk (broadcast channels, the strategy maxmzng the sum-rate

More information

IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL. 66, NO. 10, OCTOBER

IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL. 66, NO. 10, OCTOBER IEEE TRANSACTIONS ON VEICULAR TECNOLOGY, VOL. 66, NO. 10, OCTOBER 2017 8979 Robust MMSE Transcever Desgn for Nonregeneratve Multcastng MIMO Relay Systems Lenn Gopal, Member, IEEE, Yue Rong, Senor Member,

More information

The Order Relation and Trace Inequalities for. Hermitian Operators

The Order Relation and Trace Inequalities for. Hermitian Operators Internatonal Mathematcal Forum, Vol 3, 08, no, 507-57 HIKARI Ltd, wwwm-hkarcom https://doorg/0988/mf088055 The Order Relaton and Trace Inequaltes for Hermtan Operators Y Huang School of Informaton Scence

More information

MMA and GCMMA two methods for nonlinear optimization

MMA and GCMMA two methods for nonlinear optimization MMA and GCMMA two methods for nonlnear optmzaton Krster Svanberg Optmzaton and Systems Theory, KTH, Stockholm, Sweden. krlle@math.kth.se Ths note descrbes the algorthms used n the author s 2007 mplementatons

More information

LOW BIAS INTEGRATED PATH ESTIMATORS. James M. Calvin

LOW BIAS INTEGRATED PATH ESTIMATORS. James M. Calvin Proceedngs of the 007 Wnter Smulaton Conference S G Henderson, B Bller, M-H Hseh, J Shortle, J D Tew, and R R Barton, eds LOW BIAS INTEGRATED PATH ESTIMATORS James M Calvn Department of Computer Scence

More information

Robust transceiver design for AF MIMO relay systems with column correlations

Robust transceiver design for AF MIMO relay systems with column correlations Ttle Robust transcever desgn for AF MIMO relay systems wth column correlatons Author(s) Xng, C; Fe, Z; Wu, YC; Ma, S; Kuang, J Ctaton The 0 IEEE Internatonal Conference on Sgnal rocessng, Communcatons

More information

The Prncpal Component Transform The Prncpal Component Transform s also called Karhunen-Loeve Transform (KLT, Hotellng Transform, oregenvector Transfor

The Prncpal Component Transform The Prncpal Component Transform s also called Karhunen-Loeve Transform (KLT, Hotellng Transform, oregenvector Transfor Prncpal Component Transform Multvarate Random Sgnals A real tme sgnal x(t can be consdered as a random process and ts samples x m (m =0; ;N, 1 a random vector: The mean vector of X s X =[x0; ;x N,1] T

More information

The Minimum Universal Cost Flow in an Infeasible Flow Network

The Minimum Universal Cost Flow in an Infeasible Flow Network Journal of Scences, Islamc Republc of Iran 17(2): 175-180 (2006) Unversty of Tehran, ISSN 1016-1104 http://jscencesutacr The Mnmum Unversal Cost Flow n an Infeasble Flow Network H Saleh Fathabad * M Bagheran

More information

Module 3 LOSSY IMAGE COMPRESSION SYSTEMS. Version 2 ECE IIT, Kharagpur

Module 3 LOSSY IMAGE COMPRESSION SYSTEMS. Version 2 ECE IIT, Kharagpur Module 3 LOSSY IMAGE COMPRESSION SYSTEMS Verson ECE IIT, Kharagpur Lesson 6 Theory of Quantzaton Verson ECE IIT, Kharagpur Instructonal Objectves At the end of ths lesson, the students should be able to:

More information

Power Allocation/Beamforming for DF MIMO Two-Way Relaying: Relay and Network Optimization

Power Allocation/Beamforming for DF MIMO Two-Way Relaying: Relay and Network Optimization Power Allocaton/Beamformng for DF MIMO Two-Way Relayng: Relay and Network Optmzaton Je Gao, Janshu Zhang, Sergy A. Vorobyov, Ha Jang, and Martn Haardt Dept. of Electrcal & Computer Engneerng, Unversty

More information

Multigradient for Neural Networks for Equalizers 1

Multigradient for Neural Networks for Equalizers 1 Multgradent for Neural Netorks for Equalzers 1 Chulhee ee, Jnook Go and Heeyoung Km Department of Electrcal and Electronc Engneerng Yonse Unversty 134 Shnchon-Dong, Seodaemun-Ku, Seoul 1-749, Korea ABSTRACT

More information

Parameter Estimation for Dynamic System using Unscented Kalman filter

Parameter Estimation for Dynamic System using Unscented Kalman filter Parameter Estmaton for Dynamc System usng Unscented Kalman flter Jhoon Seung 1,a, Amr Atya F. 2,b, Alexander G.Parlos 3,c, and Klto Chong 1,4,d* 1 Dvson of Electroncs Engneerng, Chonbuk Natonal Unversty,

More information

Composite Hypotheses testing

Composite Hypotheses testing Composte ypotheses testng In many hypothess testng problems there are many possble dstrbutons that can occur under each of the hypotheses. The output of the source s a set of parameters (ponts n a parameter

More information

OPTIMAL COMBINATION OF FOURTH ORDER STATISTICS FOR NON-CIRCULAR SOURCE SEPARATION. Christophe De Luigi and Eric Moreau

OPTIMAL COMBINATION OF FOURTH ORDER STATISTICS FOR NON-CIRCULAR SOURCE SEPARATION. Christophe De Luigi and Eric Moreau OPTIMAL COMBINATION OF FOURTH ORDER STATISTICS FOR NON-CIRCULAR SOURCE SEPARATION Chrstophe De Lug and Erc Moreau Unversty of Toulon LSEET UMR CNRS 607 av. G. Pompdou BP56 F-8362 La Valette du Var Cedex

More information

Econ107 Applied Econometrics Topic 3: Classical Model (Studenmund, Chapter 4)

Econ107 Applied Econometrics Topic 3: Classical Model (Studenmund, Chapter 4) I. Classcal Assumptons Econ7 Appled Econometrcs Topc 3: Classcal Model (Studenmund, Chapter 4) We have defned OLS and studed some algebrac propertes of OLS. In ths topc we wll study statstcal propertes

More information

Kernel Methods and SVMs Extension

Kernel Methods and SVMs Extension Kernel Methods and SVMs Extenson The purpose of ths document s to revew materal covered n Machne Learnng 1 Supervsed Learnng regardng support vector machnes (SVMs). Ths document also provdes a general

More information

EEE 241: Linear Systems

EEE 241: Linear Systems EEE : Lnear Systems Summary #: Backpropagaton BACKPROPAGATION The perceptron rule as well as the Wdrow Hoff learnng were desgned to tran sngle layer networks. They suffer from the same dsadvantage: they

More information

U.C. Berkeley CS294: Beyond Worst-Case Analysis Luca Trevisan September 5, 2017

U.C. Berkeley CS294: Beyond Worst-Case Analysis Luca Trevisan September 5, 2017 U.C. Berkeley CS94: Beyond Worst-Case Analyss Handout 4s Luca Trevsan September 5, 07 Summary of Lecture 4 In whch we ntroduce semdefnte programmng and apply t to Max Cut. Semdefnte Programmng Recall that

More information

Linear Approximation with Regularization and Moving Least Squares

Linear Approximation with Regularization and Moving Least Squares Lnear Approxmaton wth Regularzaton and Movng Least Squares Igor Grešovn May 007 Revson 4.6 (Revson : March 004). 5 4 3 0.5 3 3.5 4 Contents: Lnear Fttng...4. Weghted Least Squares n Functon Approxmaton...

More information

Lecture 10 Support Vector Machines II

Lecture 10 Support Vector Machines II Lecture 10 Support Vector Machnes II 22 February 2016 Taylor B. Arnold Yale Statstcs STAT 365/665 1/28 Notes: Problem 3 s posted and due ths upcomng Frday There was an early bug n the fake-test data; fxed

More information

Chapter Newton s Method

Chapter Newton s Method Chapter 9. Newton s Method After readng ths chapter, you should be able to:. Understand how Newton s method s dfferent from the Golden Secton Search method. Understand how Newton s method works 3. Solve

More information

The lower and upper bounds on Perron root of nonnegative irreducible matrices

The lower and upper bounds on Perron root of nonnegative irreducible matrices Journal of Computatonal Appled Mathematcs 217 (2008) 259 267 wwwelsevercom/locate/cam The lower upper bounds on Perron root of nonnegatve rreducble matrces Guang-Xn Huang a,, Feng Yn b,keguo a a College

More information

The Concept of Beamforming

The Concept of Beamforming ELG513 Smart Antennas S.Loyka he Concept of Beamformng Generc representaton of the array output sgnal, 1 where w y N 1 * = 1 = w x = w x (4.1) complex weghts, control the array pattern; y and x - narrowband

More information

Lecture Notes on Linear Regression

Lecture Notes on Linear Regression Lecture Notes on Lnear Regresson Feng L fl@sdueducn Shandong Unversty, Chna Lnear Regresson Problem In regresson problem, we am at predct a contnuous target value gven an nput feature vector We assume

More information

The Expectation-Maximization Algorithm

The Expectation-Maximization Algorithm The Expectaton-Maxmaton Algorthm Charles Elan elan@cs.ucsd.edu November 16, 2007 Ths chapter explans the EM algorthm at multple levels of generalty. Secton 1 gves the standard hgh-level verson of the algorthm.

More information

COMPARISON OF SOME RELIABILITY CHARACTERISTICS BETWEEN REDUNDANT SYSTEMS REQUIRING SUPPORTING UNITS FOR THEIR OPERATIONS

COMPARISON OF SOME RELIABILITY CHARACTERISTICS BETWEEN REDUNDANT SYSTEMS REQUIRING SUPPORTING UNITS FOR THEIR OPERATIONS Avalable onlne at http://sck.org J. Math. Comput. Sc. 3 (3), No., 6-3 ISSN: 97-537 COMPARISON OF SOME RELIABILITY CHARACTERISTICS BETWEEN REDUNDANT SYSTEMS REQUIRING SUPPORTING UNITS FOR THEIR OPERATIONS

More information

Chapter 5. Solution of System of Linear Equations. Module No. 6. Solution of Inconsistent and Ill Conditioned Systems

Chapter 5. Solution of System of Linear Equations. Module No. 6. Solution of Inconsistent and Ill Conditioned Systems Numercal Analyss by Dr. Anta Pal Assstant Professor Department of Mathematcs Natonal Insttute of Technology Durgapur Durgapur-713209 emal: anta.bue@gmal.com 1 . Chapter 5 Soluton of System of Lnear Equatons

More information

Appendix for Causal Interaction in Factorial Experiments: Application to Conjoint Analysis

Appendix for Causal Interaction in Factorial Experiments: Application to Conjoint Analysis A Appendx for Causal Interacton n Factoral Experments: Applcaton to Conjont Analyss Mathematcal Appendx: Proofs of Theorems A. Lemmas Below, we descrbe all the lemmas, whch are used to prove the man theorems

More information

Supporting Information

Supporting Information Supportng Informaton The neural network f n Eq. 1 s gven by: f x l = ReLU W atom x l + b atom, 2 where ReLU s the element-wse rectfed lnear unt, 21.e., ReLUx = max0, x, W atom R d d s the weght matrx to

More information

MIMO Systems and Channel Capacity

MIMO Systems and Channel Capacity MIMO Systems and Channel Capacty Consder a MIMO system wth m Tx and n Rx antennas. x y = Hx ξ Tx H Rx The power constrant: the total Tx power s x = P t. Component-wse representaton of the system model,

More information

On the Multicriteria Integer Network Flow Problem

On the Multicriteria Integer Network Flow Problem BULGARIAN ACADEMY OF SCIENCES CYBERNETICS AND INFORMATION TECHNOLOGIES Volume 5, No 2 Sofa 2005 On the Multcrtera Integer Network Flow Problem Vassl Vasslev, Marana Nkolova, Maryana Vassleva Insttute of

More information

VQ widely used in coding speech, image, and video

VQ widely used in coding speech, image, and video at Scalar quantzers are specal cases of vector quantzers (VQ): they are constraned to look at one sample at a tme (memoryless) VQ does not have such constrant better RD perfomance expected Source codng

More information

IN cooperative communication systems, relay nodes can

IN cooperative communication systems, relay nodes can IEEE TRANSACTIONS ON WIRELESS COMMUNICATIONS, VOL 4, NO 0, OCTOBER 05 553 Tomlnson-Harashma Precodng Based Transcever Desgn for MIMO Relay Systems Wth Channel Covarance Informaton Lenn opal, Member, IEEE,

More information

Time-Varying Systems and Computations Lecture 6

Time-Varying Systems and Computations Lecture 6 Tme-Varyng Systems and Computatons Lecture 6 Klaus Depold 14. Januar 2014 The Kalman Flter The Kalman estmaton flter attempts to estmate the actual state of an unknown dscrete dynamcal system, gven nosy

More information

Problem Set 9 Solutions

Problem Set 9 Solutions Desgn and Analyss of Algorthms May 4, 2015 Massachusetts Insttute of Technology 6.046J/18.410J Profs. Erk Demane, Srn Devadas, and Nancy Lynch Problem Set 9 Solutons Problem Set 9 Solutons Ths problem

More information

College of Computer & Information Science Fall 2009 Northeastern University 20 October 2009

College of Computer & Information Science Fall 2009 Northeastern University 20 October 2009 College of Computer & Informaton Scence Fall 2009 Northeastern Unversty 20 October 2009 CS7880: Algorthmc Power Tools Scrbe: Jan Wen and Laura Poplawsk Lecture Outlne: Prmal-dual schema Network Desgn:

More information

Using T.O.M to Estimate Parameter of distributions that have not Single Exponential Family

Using T.O.M to Estimate Parameter of distributions that have not Single Exponential Family IOSR Journal of Mathematcs IOSR-JM) ISSN: 2278-5728. Volume 3, Issue 3 Sep-Oct. 202), PP 44-48 www.osrjournals.org Usng T.O.M to Estmate Parameter of dstrbutons that have not Sngle Exponental Famly Jubran

More information

CSci 6974 and ECSE 6966 Math. Tech. for Vision, Graphics and Robotics Lecture 21, April 17, 2006 Estimating A Plane Homography

CSci 6974 and ECSE 6966 Math. Tech. for Vision, Graphics and Robotics Lecture 21, April 17, 2006 Estimating A Plane Homography CSc 6974 and ECSE 6966 Math. Tech. for Vson, Graphcs and Robotcs Lecture 21, Aprl 17, 2006 Estmatng A Plane Homography Overvew We contnue wth a dscusson of the major ssues, usng estmaton of plane projectve

More information

Power Allocation for Distributed BLUE Estimation with Full and Limited Feedback of CSI

Power Allocation for Distributed BLUE Estimation with Full and Limited Feedback of CSI Power Allocaton for Dstrbuted BLUE Estmaton wth Full and Lmted Feedback of CSI Mohammad Fanae, Matthew C. Valent, and Natala A. Schmd Lane Department of Computer Scence and Electrcal Engneerng West Vrgna

More information

x = x 1 + :::+ x K and the nput covarance matrces are of the form ± = E[x x y ]. 3.2 Dualty Next, we ntroduce the concept of dualty wth the followng t

x = x 1 + :::+ x K and the nput covarance matrces are of the form ± = E[x x y ]. 3.2 Dualty Next, we ntroduce the concept of dualty wth the followng t Sum Power Iteratve Water-fllng for Mult-Antenna Gaussan Broadcast Channels N. Jndal, S. Jafar, S. Vshwanath and A. Goldsmth Dept. of Electrcal Engg. Stanford Unversty, CA, 94305 emal: njndal,syed,srram,andrea@wsl.stanford.edu

More information

Comparison of the Population Variance Estimators. of 2-Parameter Exponential Distribution Based on. Multiple Criteria Decision Making Method

Comparison of the Population Variance Estimators. of 2-Parameter Exponential Distribution Based on. Multiple Criteria Decision Making Method Appled Mathematcal Scences, Vol. 7, 0, no. 47, 07-0 HIARI Ltd, www.m-hkar.com Comparson of the Populaton Varance Estmators of -Parameter Exponental Dstrbuton Based on Multple Crtera Decson Makng Method

More information

SIO 224. m(r) =(ρ(r),k s (r),µ(r))

SIO 224. m(r) =(ρ(r),k s (r),µ(r)) SIO 224 1. A bref look at resoluton analyss Here s some background for the Masters and Gubbns resoluton paper. Global Earth models are usually found teratvely by assumng a startng model and fndng small

More information

A Fast Computer Aided Design Method for Filters

A Fast Computer Aided Design Method for Filters 2017 Asa-Pacfc Engneerng and Technology Conference (APETC 2017) ISBN: 978-1-60595-443-1 A Fast Computer Aded Desgn Method for Flters Gang L ABSTRACT *Ths paper presents a fast computer aded desgn method

More information

FUZZY GOAL PROGRAMMING VS ORDINARY FUZZY PROGRAMMING APPROACH FOR MULTI OBJECTIVE PROGRAMMING PROBLEM

FUZZY GOAL PROGRAMMING VS ORDINARY FUZZY PROGRAMMING APPROACH FOR MULTI OBJECTIVE PROGRAMMING PROBLEM Internatonal Conference on Ceramcs, Bkaner, Inda Internatonal Journal of Modern Physcs: Conference Seres Vol. 22 (2013) 757 761 World Scentfc Publshng Company DOI: 10.1142/S2010194513010982 FUZZY GOAL

More information

LINEAR REGRESSION ANALYSIS. MODULE IX Lecture Multicollinearity

LINEAR REGRESSION ANALYSIS. MODULE IX Lecture Multicollinearity LINEAR REGRESSION ANALYSIS MODULE IX Lecture - 31 Multcollnearty Dr. Shalabh Department of Mathematcs and Statstcs Indan Insttute of Technology Kanpur 6. Rdge regresson The OLSE s the best lnear unbased

More information

The Study of Teaching-learning-based Optimization Algorithm

The Study of Teaching-learning-based Optimization Algorithm Advanced Scence and Technology Letters Vol. (AST 06), pp.05- http://dx.do.org/0.57/astl.06. The Study of Teachng-learnng-based Optmzaton Algorthm u Sun, Yan fu, Lele Kong, Haolang Q,, Helongang Insttute

More information

princeton univ. F 17 cos 521: Advanced Algorithm Design Lecture 7: LP Duality Lecturer: Matt Weinberg

princeton univ. F 17 cos 521: Advanced Algorithm Design Lecture 7: LP Duality Lecturer: Matt Weinberg prnceton unv. F 17 cos 521: Advanced Algorthm Desgn Lecture 7: LP Dualty Lecturer: Matt Wenberg Scrbe: LP Dualty s an extremely useful tool for analyzng structural propertes of lnear programs. Whle there

More information

For now, let us focus on a specific model of neurons. These are simplified from reality but can achieve remarkable results.

For now, let us focus on a specific model of neurons. These are simplified from reality but can achieve remarkable results. Neural Networks : Dervaton compled by Alvn Wan from Professor Jtendra Malk s lecture Ths type of computaton s called deep learnng and s the most popular method for many problems, such as computer vson

More information

2E Pattern Recognition Solutions to Introduction to Pattern Recognition, Chapter 2: Bayesian pattern classification

2E Pattern Recognition Solutions to Introduction to Pattern Recognition, Chapter 2: Bayesian pattern classification E395 - Pattern Recognton Solutons to Introducton to Pattern Recognton, Chapter : Bayesan pattern classfcaton Preface Ths document s a soluton manual for selected exercses from Introducton to Pattern Recognton

More information

Generalized Linear Methods

Generalized Linear Methods Generalzed Lnear Methods 1 Introducton In the Ensemble Methods the general dea s that usng a combnaton of several weak learner one could make a better learner. More formally, assume that we have a set

More information

Comparison of Wiener Filter solution by SVD with decompositions QR and QLP

Comparison of Wiener Filter solution by SVD with decompositions QR and QLP Proceedngs of the 6th WSEAS Int Conf on Artfcal Intellgence, Knowledge Engneerng and Data Bases, Corfu Island, Greece, February 6-9, 007 7 Comparson of Wener Flter soluton by SVD wth decompostons QR and

More information

Chapter 7 Channel Capacity and Coding

Chapter 7 Channel Capacity and Coding Wreless Informaton Transmsson System Lab. Chapter 7 Channel Capacty and Codng Insttute of Communcatons Engneerng atonal Sun Yat-sen Unversty Contents 7. Channel models and channel capacty 7.. Channel models

More information

Solutions to exam in SF1811 Optimization, Jan 14, 2015

Solutions to exam in SF1811 Optimization, Jan 14, 2015 Solutons to exam n SF8 Optmzaton, Jan 4, 25 3 3 O------O -4 \ / \ / The network: \/ where all lnks go from left to rght. /\ / \ / \ 6 O------O -5 2 4.(a) Let x = ( x 3, x 4, x 23, x 24 ) T, where the varable

More information

Lossy Compression. Compromise accuracy of reconstruction for increased compression.

Lossy Compression. Compromise accuracy of reconstruction for increased compression. Lossy Compresson Compromse accuracy of reconstructon for ncreased compresson. The reconstructon s usually vsbly ndstngushable from the orgnal mage. Typcally, one can get up to 0:1 compresson wth almost

More information

University of Alberta. Library Release Form. Title of Thesis: Joint Bandwidth and Power Allocation in Wireless Communication Networks

University of Alberta. Library Release Form. Title of Thesis: Joint Bandwidth and Power Allocation in Wireless Communication Networks Unversty of Alberta Lbrary Release Form Name of Author: Xaowen Gong Ttle of Thess: Jont Bandwdth and Power Allocaton n Wreless Communcaton Networks Degree: Master of Scence Year ths Degree Granted: 2010

More information

Unified Subspace Analysis for Face Recognition

Unified Subspace Analysis for Face Recognition Unfed Subspace Analyss for Face Recognton Xaogang Wang and Xaoou Tang Department of Informaton Engneerng The Chnese Unversty of Hong Kong Shatn, Hong Kong {xgwang, xtang}@e.cuhk.edu.hk Abstract PCA, LDA

More information

A Comparative Study for Estimation Parameters in Panel Data Model

A Comparative Study for Estimation Parameters in Panel Data Model A Comparatve Study for Estmaton Parameters n Panel Data Model Ahmed H. Youssef and Mohamed R. Abonazel hs paper examnes the panel data models when the regresson coeffcents are fxed random and mxed and

More information

Convexity preserving interpolation by splines of arbitrary degree

Convexity preserving interpolation by splines of arbitrary degree Computer Scence Journal of Moldova, vol.18, no.1(52), 2010 Convexty preservng nterpolaton by splnes of arbtrary degree Igor Verlan Abstract In the present paper an algorthm of C 2 nterpolaton of dscrete

More information

Linear dispersion code with an orthogonal row structure for simplifying sphere decoding

Linear dispersion code with an orthogonal row structure for simplifying sphere decoding tle Lnear dsperson code wth an orthogonal row structure for smplfyng sphere decodng Author(s) Da XG; Cheung SW; Yuk I Ctaton he 0th IEEE Internatonal Symposum On Personal Indoor and Moble Rado Communcatons

More information

Application of Nonbinary LDPC Codes for Communication over Fading Channels Using Higher Order Modulations

Application of Nonbinary LDPC Codes for Communication over Fading Channels Using Higher Order Modulations Applcaton of Nonbnary LDPC Codes for Communcaton over Fadng Channels Usng Hgher Order Modulatons Rong-Hu Peng and Rong-Rong Chen Department of Electrcal and Computer Engneerng Unversty of Utah Ths work

More information

The Geometry of Logit and Probit

The Geometry of Logit and Probit The Geometry of Logt and Probt Ths short note s meant as a supplement to Chapters and 3 of Spatal Models of Parlamentary Votng and the notaton and reference to fgures n the text below s to those two chapters.

More information

Chapter 11: Simple Linear Regression and Correlation

Chapter 11: Simple Linear Regression and Correlation Chapter 11: Smple Lnear Regresson and Correlaton 11-1 Emprcal Models 11-2 Smple Lnear Regresson 11-3 Propertes of the Least Squares Estmators 11-4 Hypothess Test n Smple Lnear Regresson 11-4.1 Use of t-tests

More information

Scroll Generation with Inductorless Chua s Circuit and Wien Bridge Oscillator

Scroll Generation with Inductorless Chua s Circuit and Wien Bridge Oscillator Latest Trends on Crcuts, Systems and Sgnals Scroll Generaton wth Inductorless Chua s Crcut and Wen Brdge Oscllator Watcharn Jantanate, Peter A. Chayasena, and Sarawut Sutorn * Abstract An nductorless Chua

More information

LINEAR REGRESSION ANALYSIS. MODULE IX Lecture Multicollinearity

LINEAR REGRESSION ANALYSIS. MODULE IX Lecture Multicollinearity LINEAR REGRESSION ANALYSIS MODULE IX Lecture - 30 Multcollnearty Dr. Shalabh Department of Mathematcs and Statstcs Indan Insttute of Technology Kanpur 2 Remedes for multcollnearty Varous technques have

More information

Negative Binomial Regression

Negative Binomial Regression STATGRAPHICS Rev. 9/16/2013 Negatve Bnomal Regresson Summary... 1 Data Input... 3 Statstcal Model... 3 Analyss Summary... 4 Analyss Optons... 7 Plot of Ftted Model... 8 Observed Versus Predcted... 10 Predctons...

More information

Power law and dimension of the maximum value for belief distribution with the max Deng entropy

Power law and dimension of the maximum value for belief distribution with the max Deng entropy Power law and dmenson of the maxmum value for belef dstrbuton wth the max Deng entropy Bngy Kang a, a College of Informaton Engneerng, Northwest A&F Unversty, Yanglng, Shaanx, 712100, Chna. Abstract Deng

More information

Natural Language Processing and Information Retrieval

Natural Language Processing and Information Retrieval Natural Language Processng and Informaton Retreval Support Vector Machnes Alessandro Moschtt Department of nformaton and communcaton technology Unversty of Trento Emal: moschtt@ds.untn.t Summary Support

More information

Speeding up Computation of Scalar Multiplication in Elliptic Curve Cryptosystem

Speeding up Computation of Scalar Multiplication in Elliptic Curve Cryptosystem H.K. Pathak et. al. / (IJCSE) Internatonal Journal on Computer Scence and Engneerng Speedng up Computaton of Scalar Multplcaton n Ellptc Curve Cryptosystem H. K. Pathak Manju Sangh S.o.S n Computer scence

More information

Hongyi Miao, College of Science, Nanjing Forestry University, Nanjing ,China. (Received 20 June 2013, accepted 11 March 2014) I)ϕ (k)

Hongyi Miao, College of Science, Nanjing Forestry University, Nanjing ,China. (Received 20 June 2013, accepted 11 March 2014) I)ϕ (k) ISSN 1749-3889 (prnt), 1749-3897 (onlne) Internatonal Journal of Nonlnear Scence Vol.17(2014) No.2,pp.188-192 Modfed Block Jacob-Davdson Method for Solvng Large Sparse Egenproblems Hongy Mao, College of

More information

Non-linear Canonical Correlation Analysis Using a RBF Network

Non-linear Canonical Correlation Analysis Using a RBF Network ESANN' proceedngs - European Smposum on Artfcal Neural Networks Bruges (Belgum), 4-6 Aprl, d-sde publ., ISBN -97--, pp. 57-5 Non-lnear Canoncal Correlaton Analss Usng a RBF Network Sukhbnder Kumar, Elane

More information

LECTURE 9 CANONICAL CORRELATION ANALYSIS

LECTURE 9 CANONICAL CORRELATION ANALYSIS LECURE 9 CANONICAL CORRELAION ANALYSIS Introducton he concept of canoncal correlaton arses when we want to quantfy the assocatons between two sets of varables. For example, suppose that the frst set of

More information

High resolution entropy stable scheme for shallow water equations

High resolution entropy stable scheme for shallow water equations Internatonal Symposum on Computers & Informatcs (ISCI 05) Hgh resoluton entropy stable scheme for shallow water equatons Xaohan Cheng,a, Yufeng Ne,b, Department of Appled Mathematcs, Northwestern Polytechncal

More information

APPROXIMATE PRICES OF BASKET AND ASIAN OPTIONS DUPONT OLIVIER. Premia 14

APPROXIMATE PRICES OF BASKET AND ASIAN OPTIONS DUPONT OLIVIER. Premia 14 APPROXIMAE PRICES OF BASKE AND ASIAN OPIONS DUPON OLIVIER Prema 14 Contents Introducton 1 1. Framewor 1 1.1. Baset optons 1.. Asan optons. Computng the prce 3. Lower bound 3.1. Closed formula for the prce

More information

A Hybrid Variational Iteration Method for Blasius Equation

A Hybrid Variational Iteration Method for Blasius Equation Avalable at http://pvamu.edu/aam Appl. Appl. Math. ISSN: 1932-9466 Vol. 10, Issue 1 (June 2015), pp. 223-229 Applcatons and Appled Mathematcs: An Internatonal Journal (AAM) A Hybrd Varatonal Iteraton Method

More information

A New Refinement of Jacobi Method for Solution of Linear System Equations AX=b

A New Refinement of Jacobi Method for Solution of Linear System Equations AX=b Int J Contemp Math Scences, Vol 3, 28, no 17, 819-827 A New Refnement of Jacob Method for Soluton of Lnear System Equatons AX=b F Naem Dafchah Department of Mathematcs, Faculty of Scences Unversty of Gulan,

More information

The Two-scale Finite Element Errors Analysis for One Class of Thermoelastic Problem in Periodic Composites

The Two-scale Finite Element Errors Analysis for One Class of Thermoelastic Problem in Periodic Composites 7 Asa-Pacfc Engneerng Technology Conference (APETC 7) ISBN: 978--6595-443- The Two-scale Fnte Element Errors Analyss for One Class of Thermoelastc Problem n Perodc Compostes Xaoun Deng Mngxang Deng ABSTRACT

More information

Dr. Shalabh Department of Mathematics and Statistics Indian Institute of Technology Kanpur

Dr. Shalabh Department of Mathematics and Statistics Indian Institute of Technology Kanpur Analyss of Varance and Desgn of Exerments-I MODULE III LECTURE - 2 EXPERIMENTAL DESIGN MODELS Dr. Shalabh Deartment of Mathematcs and Statstcs Indan Insttute of Technology Kanur 2 We consder the models

More information

Stanford University CS359G: Graph Partitioning and Expanders Handout 4 Luca Trevisan January 13, 2011

Stanford University CS359G: Graph Partitioning and Expanders Handout 4 Luca Trevisan January 13, 2011 Stanford Unversty CS359G: Graph Parttonng and Expanders Handout 4 Luca Trevsan January 3, 0 Lecture 4 In whch we prove the dffcult drecton of Cheeger s nequalty. As n the past lectures, consder an undrected

More information

Resource Allocation with a Budget Constraint for Computing Independent Tasks in the Cloud

Resource Allocation with a Budget Constraint for Computing Independent Tasks in the Cloud Resource Allocaton wth a Budget Constrant for Computng Independent Tasks n the Cloud Wemng Sh and Bo Hong School of Electrcal and Computer Engneerng Georga Insttute of Technology, USA 2nd IEEE Internatonal

More information

Feature Selection: Part 1

Feature Selection: Part 1 CSE 546: Machne Learnng Lecture 5 Feature Selecton: Part 1 Instructor: Sham Kakade 1 Regresson n the hgh dmensonal settng How do we learn when the number of features d s greater than the sample sze n?

More information

Chat eld, C. and A.J.Collins, Introduction to multivariate analysis. Chapman & Hall, 1980

Chat eld, C. and A.J.Collins, Introduction to multivariate analysis. Chapman & Hall, 1980 MT07: Multvarate Statstcal Methods Mke Tso: emal mke.tso@manchester.ac.uk Webpage for notes: http://www.maths.manchester.ac.uk/~mkt/new_teachng.htm. Introducton to multvarate data. Books Chat eld, C. and

More information

Lecture 3. Ax x i a i. i i

Lecture 3. Ax x i a i. i i 18.409 The Behavor of Algorthms n Practce 2/14/2 Lecturer: Dan Spelman Lecture 3 Scrbe: Arvnd Sankar 1 Largest sngular value In order to bound the condton number, we need an upper bound on the largest

More information

A new Approach for Solving Linear Ordinary Differential Equations

A new Approach for Solving Linear Ordinary Differential Equations , ISSN 974-57X (Onlne), ISSN 974-5718 (Prnt), Vol. ; Issue No. 1; Year 14, Copyrght 13-14 by CESER PUBLICATIONS A new Approach for Solvng Lnear Ordnary Dfferental Equatons Fawz Abdelwahd Department of

More information

Semi-supervised Classification with Active Query Selection

Semi-supervised Classification with Active Query Selection Sem-supervsed Classfcaton wth Actve Query Selecton Jao Wang and Swe Luo School of Computer and Informaton Technology, Beng Jaotong Unversty, Beng 00044, Chna Wangjao088@63.com Abstract. Labeled samples

More information

Inexact Newton Methods for Inverse Eigenvalue Problems

Inexact Newton Methods for Inverse Eigenvalue Problems Inexact Newton Methods for Inverse Egenvalue Problems Zheng-jan Ba Abstract In ths paper, we survey some of the latest development n usng nexact Newton-lke methods for solvng nverse egenvalue problems.

More information

e - c o m p a n i o n

e - c o m p a n i o n OPERATIONS RESEARCH http://dxdoorg/0287/opre007ec e - c o m p a n o n ONLY AVAILABLE IN ELECTRONIC FORM 202 INFORMS Electronc Companon Generalzed Quantty Competton for Multple Products and Loss of Effcency

More information

Outline. Bayesian Networks: Maximum Likelihood Estimation and Tree Structure Learning. Our Model and Data. Outline

Outline. Bayesian Networks: Maximum Likelihood Estimation and Tree Structure Learning. Our Model and Data. Outline Outlne Bayesan Networks: Maxmum Lkelhood Estmaton and Tree Structure Learnng Huzhen Yu janey.yu@cs.helsnk.f Dept. Computer Scence, Unv. of Helsnk Probablstc Models, Sprng, 200 Notces: I corrected a number

More information

An Upper Bound on SINR Threshold for Call Admission Control in Multiple-Class CDMA Systems with Imperfect Power-Control

An Upper Bound on SINR Threshold for Call Admission Control in Multiple-Class CDMA Systems with Imperfect Power-Control An Upper Bound on SINR Threshold for Call Admsson Control n Multple-Class CDMA Systems wth Imperfect ower-control Mahmoud El-Sayes MacDonald, Dettwler and Assocates td. (MDA) Toronto, Canada melsayes@hotmal.com

More information

An Admission Control Algorithm in Cloud Computing Systems

An Admission Control Algorithm in Cloud Computing Systems An Admsson Control Algorthm n Cloud Computng Systems Authors: Frank Yeong-Sung Ln Department of Informaton Management Natonal Tawan Unversty Tape, Tawan, R.O.C. ysln@m.ntu.edu.tw Yngje Lan Management Scence

More information

Primer on High-Order Moment Estimators

Primer on High-Order Moment Estimators Prmer on Hgh-Order Moment Estmators Ton M. Whted July 2007 The Errors-n-Varables Model We wll start wth the classcal EIV for one msmeasured regressor. The general case s n Erckson and Whted Econometrc

More information

Markov Chain Monte Carlo Lecture 6

Markov Chain Monte Carlo Lecture 6 where (x 1,..., x N ) X N, N s called the populaton sze, f(x) f (x) for at least one {1, 2,..., N}, and those dfferent from f(x) are called the tral dstrbutons n terms of mportance samplng. Dfferent ways

More information

ANSWERS. Problem 1. and the moment generating function (mgf) by. defined for any real t. Use this to show that E( U) var( U)

ANSWERS. Problem 1. and the moment generating function (mgf) by. defined for any real t. Use this to show that E( U) var( U) Econ 413 Exam 13 H ANSWERS Settet er nndelt 9 deloppgaver, A,B,C, som alle anbefales å telle lkt for å gøre det ltt lettere å stå. Svar er gtt . Unfortunately, there s a prntng error n the hnt of

More information

Error Probability for M Signals

Error Probability for M Signals Chapter 3 rror Probablty for M Sgnals In ths chapter we dscuss the error probablty n decdng whch of M sgnals was transmtted over an arbtrary channel. We assume the sgnals are represented by a set of orthonormal

More information

Cognitive Access Algorithms For Multiple Access Channels

Cognitive Access Algorithms For Multiple Access Channels 203 IEEE 4th Workshop on Sgnal Processng Advances n Wreless Communcatons SPAWC) Cogntve Access Algorthms For Multple Access Channels Ychuan Hu and Alejandro Rbero, Department of Electrcal and Systems Engneerng,

More information

Queueing Networks II Network Performance

Queueing Networks II Network Performance Queueng Networks II Network Performance Davd Tpper Assocate Professor Graduate Telecommuncatons and Networkng Program Unversty of Pttsburgh Sldes 6 Networks of Queues Many communcaton systems must be modeled

More information

1 Convex Optimization

1 Convex Optimization Convex Optmzaton We wll consder convex optmzaton problems. Namely, mnmzaton problems where the objectve s convex (we assume no constrants for now). Such problems often arse n machne learnng. For example,

More information

Lecture 20: November 7

Lecture 20: November 7 0-725/36-725: Convex Optmzaton Fall 205 Lecturer: Ryan Tbshran Lecture 20: November 7 Scrbes: Varsha Chnnaobreddy, Joon Sk Km, Lngyao Zhang Note: LaTeX template courtesy of UC Berkeley EECS dept. Dsclamer:

More information

Parametric fractional imputation for missing data analysis. Jae Kwang Kim Survey Working Group Seminar March 29, 2010

Parametric fractional imputation for missing data analysis. Jae Kwang Kim Survey Working Group Seminar March 29, 2010 Parametrc fractonal mputaton for mssng data analyss Jae Kwang Km Survey Workng Group Semnar March 29, 2010 1 Outlne Introducton Proposed method Fractonal mputaton Approxmaton Varance estmaton Multple mputaton

More information

Temperature. Chapter Heat Engine

Temperature. Chapter Heat Engine Chapter 3 Temperature In prevous chapters of these notes we ntroduced the Prncple of Maxmum ntropy as a technque for estmatng probablty dstrbutons consstent wth constrants. In Chapter 9 we dscussed the

More information