On Power Allocation for Distributed Detection with Correlated Observations and Linear Fusion

Size: px
Start display at page:

Download "On Power Allocation for Distributed Detection with Correlated Observations and Linear Fusion"

Transcription

1 1 On Power Alloation for Distributed Detetion with Correlated Observations and Linear Fusion Hamid R Ahmadi, Member, IEEE, Nahal Maleki, Student Member, IEEE, Azadeh Vosoughi, Senior Member, IEEE arxiv:1794v1 [eesssp] 6 Ot 17 Abstrat We onsider a binary hypothesis testing problem in an inhomogeneous wireless sensor network, where a fusion enter (FC) makes a global deision on the underlying hypothesis We assume sensors observations are orrelated Gaussian and sensors are unaware of this orrelation when making deisions Sensors send their modulated deisions over fading hannels, subjet to individual and/or total transmit power onstraints For parallel-aess hannel (PAC) and multiple-aess hannel (MAC) models, we derive modified defletion oeffiient (MDC) of the test statisti at the FC with oherent reeption We propose ransmit power alloation sheme, whih maximizes MDC of the test statisti, under three different sets of transmit power onstraints: total power onstraint, individual and total power onstraints, individual power onstraints only When analytial solutions to our onstrained optimization problems are elusive, we disuss how these problems an be onverted to onvex ones We study how orrelation among sensors observations, reliability of loal deisions, ommuniation hannel model and hannel qualities and transmit power onstraints affet the reliability of the global deision and power alloation of inhomogeneous sensors Index Terms Distributed detetion, oherent reeption, modified defletion oeffiient, power alloation, orrelated observations, linear fusion, parallel-aess hannel, multiple-aess hannel I INTRODUCTION The lassial problem of binary distributed detetion in a network onsisting of multiple distributed sensors and a fusion enter (FC), has a long and rih history Eah sensor (loal detetor) proesses its single observation loally and passes its binary deision to the FC, that is tasked with fusing the binary deisions reeived from the individual sensors and deiding whih of the two underlying hypotheses is true [1] [3] Motivated by the potential appliation of wireless sensor networks (WSNs) for event monitoring, researhers have further studied this problem and extended its setup, taking into aount that bandwidth-onstrained ommuniation hannels between sensors and the FC are error-prone, due to limited transmit power to ombat noise and fading (so-alled hannel aware binary distributed detetion [4] [6]) Given eah sensor makes its binary deision based on one loal observation, they have investigated how the reliability of the final deision at the FC is affeted by performane indies of loal detetors (sensors) as well as wireless hannel properties Following these works, we onsider hannel aware binary distributed detetion in a WSN with oherent reeption at the FC [7] [9] In this paper, our goal is to study transmit power alloation, when eah sensor has an individual transmit power onstraint and/or all sensors have a joint transmit power onstraint, suh that the reliability of the final deision at the FC is maximized This work is supported by the National Siene Foundation under grants CCF and CCF Power alloation for hannel aware binary distributed detetion in WSNs has been studied in [], [11] More speifially, [] studied the power alloation that maximizes the J-divergene between the distributions of the reeived signals at the FC under two different hypotheses, subjet to individual and total transmit power onstraints on the sensors, with parallel aess hannel (PAC) 1 and oherent reeption at the FC (ie, hannel phases are known and ompensated at the sensors) Leveraging on [], [11] studied detetion outage and detetion diversity, as the number of sensors goes to infinity, and sensors have idential performane indies Note that [], [11] assume the sensors have unorrelated observations under eah hypothesis Power alloation in WSNs has also been studied for distributed estimation [13] [], where some works minimized the mean square error (MSE) of an estimator subjet to ertain transmit power onstraints [14] [1], while others minimized total transmit power subjet to a onstraint on the MSE of an estimator [13], [] These works, exept [13], [19], [], mainly fous on PAC with oherent reeption at the FC In [13], [19], [], sensors and the FC are onneted differently via a multiple-aess hannel (MAC), where the individual sensors send their signals simultaneously, albeit after hannel phases are ompensated at the sensors, and the FC reeives the oherent sum of these transmitted signals Most of these works assume the sensors observations are unorrelated, with the exeption of [14], [16], [] In [19], [] sensors ollaborate with eah other by linearly ombining their independent observations before sending to the FC For binary distributed detetion in WSNs, [1] ompared the detetion performane using both PAC and MAC, with linear fusion rule and nonoherent reeption at the FC (ie, no hannel phase ompensation at the sensors), albeit without imposing any transmit power onstraint Assuming the sensors observations are unorrelated under eah hypothesis and the FC utilizes a linear fusion rule when using PAC, [1] showed that oherent MAC outperforms oherent PAC, whereas nonoherent PAC (MAC) outperforms nonoherent MAC (PAC) when sensors deisions are (un)reliable Distributed detetion with orrelated observations has been studied assuming errorfree [3], [3], [4] and erroneous ommuniation hannels [] The fous of these works though is on how to design optimal loal and global deisions rules to improve the detetion reliability at the FC, assuming sensors know the orrelation among their observations Different from [3], [3] [] we fous on how to optimally transmit the sensors deisions to 1 In PAC, hannels between the sensors and the FC are orthogonal (noninterfering) This an be realized by either time, frequeny, or ode division multiple aess [1], [13]

2 the FC within ertain transmit power onstraints, with a linear fusion rule at the FC and assuming sensors are unaware of the orrelation among their observations Our Contributions: We onsider a binary hypothesis testing problem using M sensors and a FC, where under H, sensors observations are unorrelated Gaussian with ovariane matrix σi and under H 1 they are orrelated Gaussian with a non-diagonal ovariane matrix Σ We relax the assumption in [3], [3] [] that sensors know the orrelation among their observations and onsider a more pratial senario, where the sensors are unaware of suh orrelation Sensors send their modulated binary deisions over nonideal fading hannels, subjet to individual and/or total transmit power onstraints We onsider PAC and MAC with oherent reeption at the FC, assuming that hannel phases are ompensated at the sensors similar to [13], [19], [] To urb the hardware and omputational omplexity and also have a fair omparison between PAC and MAC, we assume that, when the sensors and the FC are onneted via PAC, the FC utilizes a linear fusion rule to obtain the global test statisti T We propose a transmit power alloation sheme, whih maximizes modified defletion oeffiient (MDC) of T We hoose MDC as the performane metri, sine unlike detetion probability and J- divergene that require the probability distribution funtion of T, obtaining MDC only needs the first and seond order statistis of T, and often renders a losed-form expression [6] [8] Also, an MDC-based optimization problem an lead into near-optimal solutions for its orresponding detetion probability-based optimization problem with muh less omputational omplexity [7], [6], [9] We obtain the MDC of T for oherent PAC and MAC in losed-forms that depends on the orrelation among sensors observations Considering three different sets of transmit power onstraints, we investigate transmit power alloation shemes that maximize the MDC Under the onditions that analytial solutions to our onstrained optimization problems are elusive, we disuss how these problems an be onverted to onvex ones and thus an be solved numerially Paper Organization: Setion II details our system model and three different sets of transmit power onstraints Setion III derives the MDC of T for oherent PAC and MAC in losed-form expressions Setion IV formulates three different sets of onstrained optimization problems and desribes our approah to solve these problems Setion V presents our numerial results for different orrelation values, sensors observations and ommuniation hannel qualities Setion VI onludes the paper Notations: Salars, vetors and matries are denoted by non-boldfae lower, boldfae lower, and boldfae upper ase letters, respetively A Gaussian random vetor x with mean vetor µ and ovariane matrix Σ is shown as x N (µ, Σ) Transpose and omplex onjugate transpose (Hermitian) of vetor a are denoted as a T and a H, respetively DIAG{a} represents a diagonal matrix whose diagonal elements are the omponents of olumn vetor a A (A ) indiates that A is a positive (semi-)definite matrix a b (a b) indiates that eah entry of a is greater than (or equal to) the orresponding entry of b Re{x} is the real part of x = [,, ] T and 1 = [1,, 1] T are two M 1 vetors The (i, j) entry of matrix A is indiated with [A] ij For vetor a we have a = a T a and a = a T a II SYSTEM MODEL AND PROBLEM STATEMENT Our system model onsists of an FC and M distributed sensors with observation vetor x = [x 1, x,, x M ] T The FC is tasked with solving the binary hypothesis testing problem H : x N (, σ I), H 1 : x N (, Σ), where σ is the variane under H and Σ is a non-diagonal ovariane matrix under H 1 with diagonal entries different from σ, ie, under H 1 (H ) sensors observations are orrelated (unorrelated) Gaussian variables with different energy levels Suppose sensor k, only based on its own observation x k, makes a binary deision [], [], [1] and maps it to u k = 1 (u k = ) when it deides H 1 (H ), ie, we assume that sensor k is unaware of the orrelation among sensors observations, Σ We denote p fk = P(u k = 1 H ) and p dk = P(u k = 1 H 1 ) as the false alarm and detetion probabilities of sensor k and assume p dk > p fk The deision u k is ommuniated to the FC over a fading hannel with transmit power P tk Let h k = h k e jϕ k denote the omplex fading oeffiient orresponding to sensor k, with h k and ϕ k being the hannel amplitude and phase, respetively Let y k and y denote the hannel output orresponding to the hannel input u k and (u 1, u,, u M ), when the sensors and the FC are onneted via PAC and MAC, respetively Sine hannel phases are ompensated at the sensors, we have [1] PAC : y k = P k h k u k + n k, k = 1,, M and M MAC : y = Pk h k u k + n (1) k=1 where ommuniation hannel noises are n k CN (, σn) and n CN (, σn) Fading oeffiients h k s and noises n k s and n are all mutually unorrelated and h k is assumed to be onstant during a detetion interval Also P k = P tk θ k, where θ k = Gd ɛ F S k, d F Sk is the distane between sensor k and the FC, ɛ is the pathloss exponent, and G is a onstant We assume that the FC obtains est statisti T from the hannel output(s) and makes a global deision u {, 1} where u = 1 and u = orrespond to H 1 and u =1 H, respetively In partiular, the FC applies T τ u = where the threshold τ is hosen to maximize the total detetion probability P D = P(u = 1 H 1 ) at the FC, subjet to the onstraint that the total false alarm probability satisfies P F = P(u = 1 H ) β F at the FC, where β F (, 1) In a PAC, we assume that the FC is restrited to utilize a linear fusion rule to obtain the test statisti T [], [1] Implementing the linear fusion rule has low omplexity and allows a fair omparison between PAC and MAC Furthermore, the authors in [] have shown that, when idential sensors and the FC are onneted via PAC, the linear fusion rule is a good approximation to the optimal Likelihood Ratio Test (LRT) rule at low signal-to-noise-ratio (SNR) regime We let T be

3 3 PAC : T = M Re(y k ), MAC : T = Re(y) () k=1 We onsider oherent PAC and MAC with hannel phase ompensation at the sensors [], [] Our goal is to find the transmit powers at sensors suh that the MDC of T is maximized, subjet to different sets of power onstraints We refer to these as the MDC-based transmit power alloation We onsider three different sets of transmit power onstraints: (A) there is otal power onstraint (TPC) suh that M k=1 P t k P tot, where P tot is the total transmit power budget among sensors, we refer to this set as TPC; (B) there is an individual power onstraint (IPC) for eah sensor suh that P tk P k as well as a TPC M k=1 P t k P tot, where P tot < M k=1 P k, we refer to this set as TIPC; (C) there are only IPCs for sensors suh that P tk P k, we refer to this set as IPC Setion III drives the MDC of T for oherent PAC and MAC The MDC-based transmit power alloations under these three different sets of power onstraints are disussed in Setion IV III DERIVING MODIFIED DEFLECTION COEFFICIENT Before delving into the derivations, we introdue the following definitions and notations Consider the signal model in (1) and () We let a k = P k, w k =Re(n k ), w =Re(n) We define the olumn vetors h = [h 1,, h M ] T, h = [ h 1,, h M ] T, y = [y 1,, y M ] T, a = [a 1,, a M ] T, w = [w 1,, w M ] T, n = [n 1,, n M ] T, p d = [p d1,, p dm ] T, p f = [p f1,, p fm ] T, u = [u 1, u,, u M ] T, P = [P 1,, P M ] T, ψ = [ψ 1,, ψ M ], φ=[φ 1,, φ M ], and the square matrix H =DIAG{ h } We define the MDC of T as [6] ( E{T H1, h} E{T H, h} ) MDC =, (3) Var{T H 1, h} where E{} and Var{} are performed with respet to the hannel inputs u k s and the hannel noises To alulate E{T H i, h} for i =, 1 and Var{T H 1, h} in (3), we use the Bayes rule and the fat that H i u k y k (y) u in PAC(MAC) form Markov hains for i=, 1 Hene E{T H i, h} = E{T u, h}p(u H i ), i =, 1 (4) u Var{T H 1, h} = P(u H 1 ), () u where { } = E (T E{T H 1, h}) u, h, and the sums are taken over all values of vetor u To simplify in (6), we add and subtrat E{T u, h} to the terms inside the parenthesis in (6) and expand the produts We have = (6) E { (T E{T u, h}) u } + (E{T u, h} E{T H 1, h}) }{{}}{{} + E {(T E{T u, h})(e{t u, h} E{T H 1, h}) u} We observe that the last term in (6) is zero Thus in (6) is simplified to = + Using (4), (6) and (6), we derive the MDC in the following A PAC Considering the signal model in (1) and (), we have Re(y k ) = a k h k u k + w k where w k N (, σ n ) We write T =a T H u+1 T w Therefore E{T u, h}=a T H u Substituting E{T u, h} into (4) and using the fats p d =E{u H 1 }= u up(u H 1) and p f =E{u H }= u up(u H ) we find E{T H 1, h}=a T H p d, and E{T H, h}=a T H p f (7) Next, we derive, for Var{T H 1, h} Sine T E{T u, h} = 1 T w, we find = E{1 T ww T 1} = M σ n Also, beause E{T u, h} E{T H 1, h} = a T H (u p d ), we have = a T H (u p d )(u p d ) T H a Substituting = + into (6) and using the fats u P(u H 1) = 1, u (u p d)(u p d ) T P(u H 1 ) = E{uu T H 1 } p d p T d, we reah to Var{T H 1, h} = M σ n + at H ( P d p d p T d ) H a, (8) where P d = E{uu T H 1 } is a square matrix with diagonal entries [ P d ] ii = p di and off-diagonal entries [ P d ] ij = P(u i = 1, u j = 1 H 1 ) for i, j = 1,, M, i j Note that the orrelation among sensors observations affets the off-diagonal entries of Pd, ie, for independent observations [ P d ] ij =p di p dj for all i j and equivalently P d =DIAG{p d }(I DIAG{p d }) + p d p T d (9) Substituting (7), (8) into (3) we have where MDC(a) = at bb T a a T Ka + () b = H (p d p f ), = M σ n, K = H ( P d p d p T d ) H B MAC Considering the signal model in (1) and (), we have Re(y) = M k=1 a k h k u k + w where w N (, σ n ) We write T = a T H u + w Therefore E{T u, h} = a T H u Substituting E{T u, h} into (4) and applying similar fats as stated above, we find E{T H 1, h}=a T H p d, and E{T H, h}=a T H p f (11) Next, we find and Sine T E{T u, h} = w, we find = E{w } = σ n Also, sine E{T u, h} E{T H 1, h} = a T H (u p d ), we have =a T H (u p d )(u p d ) T H a Substituting = + into (6) and using similar fats as stated above we reah Var{T H 1, h} = σ n + at H ( P d p d p T d ) H a (1) Substituting (11), (1) into (3) we have MDC(a) = at bb T a a T Ka + (13)

4 4 where b = H (p d p f ), = σ n, K = H ( P d p d p T d ) H Regarding the results in () and (13), a remark follows Remark: For both PAC and MAC, the MDC takes the following form MDC(a) = at bb T a a T Ka + (14) Vetor b and matrix K are idential for PAC and MAC, whereas salar, whih aptures the effet of the hannel noises, is M times larger in PAC Note that b and K depend on the hannel amplitudes H and the loal performane indies Furthermore, K depends on the spatial orrelation among sensors observations IV MDC-BASED TRANSMIT POWER ALLOCATION Reall P k = P tk θ k where P tk is transmit power of sensor k and θ k aptures the pathloss effet Sine a k = P k, we define k = P tk = a k θk Let = [1,, M ] T, P t = [P t1,, P tm ] T, and Θ be the omponent-wise square root of Θ = DIAG{[θ 1,, θ M ] T } We an rewrite (14) expliitly in terms of vetor as MDC( ) = at t b t b T t a T t K t +, (1) where b t = Θb and K t = ΘK Θ In this setion, we maximize the MDC in (1), with respet to, subjet to different sets of power onstraints speified in Setion II: (A) TPC, where a T t P tot ; (B) TIPC, where a T t P tot and P We define vetor P = [P 1,, P M ] T and P is the omponent-wise square root of P ; (C) IPC, where P Setions IV-A, IV-B, IV-C disuss the analytial solutions for MDC-based power alloations under these different sets of power onstraints A Maximizing MDC in (1) under TPC The MDC-based transmit power alloation under TPC is the solution to the following problem max st a T t btbt t at a T t Ktat+ (O 1 ) a T t P tot We start with Lemma 1 whih states that the solution to (O 1 ) satisfies TPC at equality Lemma 1 The maximum values of MDC in (1) are ahieved when the inequality onstraint a T t P tot turns into equality onstraint Proof: Suppose 1 maximizes MDC and a T t11 <P tot Define = at1 Ptot, whih satisfies a T t =P tot We have MDC( ) = 1 a T t1 btbt t at1 a T t1 Ktat1+( at t1 1 P tot ) > at t1 btbt t at1 a T t1 Ktat1+ = MDC(1 ), whih ontradits the optimality assumption of 1 ie, the that maximizes MDC, must satisfy a T t = P tot When the inequality onstraint in TPC is turned into equality onstraint, we an rewrite MDC in (1) as MDC( )= at t b t b T t a T t Q a, where Q a =K t + P tot I (16) Hene, (O 1 ) redues to max st a T t btbt t at (O a 1) T t Qaat a T t = P tot To analytially solve (O 1), we use the result of Lemma given below Lemma For Q the funtion f(x) = xt b tb T t x x T Qx maximized at x =Q 1 b t and its non-zero sales Proof: See Appendix A To be able to use Lemma to solve (O 1), we need to examine whether symmetri matrix Q a is positive definite Note P d p d p T d sine it is a ovariane matrix Thus K, K t Also P tot I Therefore Q a To solve (O 1), we find ˆq = q q where q = Q 1 a b t If ˆq we let a t = ˆq P tot and if ˆq we let a t = ˆq P tot But if all the entries of ˆq do not have the same sign, we resort to numerial solutions In partiular, we turn the problem (O 1) into a onvex problem and solve it numerially We disuss these numerial solutions in Setion IV-D Analytial Solution to (O 1) with Independent Observations: Pd is given in (9) and K simplifies to K = H DIAG{p d }(I DIAG{p d }) H Let g k = θ k h k It is easy to verify Q a is a diagonal matrix with diagonal entries [Q a ] kk = p dk (1 p dk )gk + P tot Let q k be the kth entry of q = Q 1 (p a b t Then q k = dk p fk )g k p dk (1 p dk )gk +, k = 1,, M, P tot whih is positive for p dk >p fk We observe q k (p d k p fk ) p dk (1 p dk )g k for large Ptot, whereas q k Ptot (p d k p fk )g k for small Ptot For homogeneous sensors where p fk = p f and p dk = p d, we find the MDC-based power alloation strategy as q k 1 g k for large Ptot (inverse water filling) and q k g k for small Ptot (water filling) B Maximizing MDC in (1) under TIPC The MDC-based transmit power alloation is the solution to the following problem max st a T t btbt t at a T t Ktat+ (O ) a T t P tot P While analytial solution to (O ) remains elusive, we find sub-optimal power alloation via solving the following optimization problem max st a T t btbt t at (O a ) T t Qaat a T t = P tot P where Q a is given in (16) Note that (O ) is idential to (O ), exept that the inequality in TPC is turned into equality, ie, is

5 the feasible set of (O ) is a subset of the feasible set of (O ) and the objetive funtion of (O ) is rewritten aordingly Indeed, this sub-optimal solution is an aurate solution when κ = PtotgT g 1 for (O ), as we show in the following Examining K and K t when κ 1, we an establish the following inequalities a T (a) t K t a T t Θ H 11 T H Θ = a T t gg T (b) (a T t )(g T g) () P tot g T g (d) where (a) is obtained noting that all entries of Pd p d p T d are less that 1, (b) is found using Cauhy-Shwarz inequality, () omes from the inequality onstraint in (O ), and (d) is due to κ 1 This implies that when κ 1, (O ) an be approximated as (O l ) in (17) min (O a ) l T t btbt t at st a T t P tot (17) P In Appendix B, we show that the solution to (O) l satisfies the equality a T t = P tot This onfirms that the solution to (O ) (sub-optimal solution) is an aurate substitute for the solution to (O ) under the ondition κ 1 To solve (O ), we first ignore the box onstraints of IPC and onsider only TPC at equality The problem solving strategy is similar to solving (O 1) in Setion IV-A In partiular, to solve (O ), we find ˆq = q q where q =Q 1 a b t If ˆq we let a t1 = ˆq P tot and if ˆq we let a t1 = ˆq P tot If a t1 satisfies the box onstraint P, it is the solution to (O ) However, if a t1 does not satisfy its orresponding box onstraint, following Appendix A, we an easily show that f(x) = xt b tb T t x does x T Qx not have loal maximum or minimum in the set {x : x } This means that, in this ase, the losest feasible point to a t1 is the solution to (O ) That is, the solution to (O ) when a t1 P is the solution to (O ) given below min a t1 (O ) st a T t = P tot P Our analytial solution to (O ) is presented in the appendix C Note that (O ) is not onvex In Appendix C we show that, despite this fat, the solution to Karush-Kuhn-Tuker (KKT) onditions for (O ) is unique C Maximizing MDC in (1) under IPC The MDC-based transmit power alloation is the solution to the following optimization problem max a T t btbt t at a T t Ktat+ (O 3 ) st P Similar to Setion IV-B, we show below that, when ξ = 1 T P g T g 1, (O 3 ) an be approximated as (O3) l in (18) Examining K and K t when ξ 1, we an establish the following inequalities a T (a) t K t a T t Θ H 11 T H Θ = a T t gg T (b) (a T t a)(g T g) () 1 T P g T g (d), where (a) is beause all entries of P d p d p T d are less that 1, (b) is found using Cauhy-Shwarz inequality, () omes from the inequality in IPC, and (d) is due to ξ 1 This implies that when ξ 1, (O 3 ) an be approximated with (O3) l in (18) min a T t btbt t at (O l 3) st P (18) In Appendix D we show that the solution to (O3) l is = P Analytial Solution to (O 3 ) with Independent Observations: Suppose P k = P, k = 1,, M We showed in Setion IV-A that K, K t With independent observations, these matries beome diagonal To solve (O 3 ), we minimize 1 under IPC Assume ψ, φ are the Lagrange multi- MDC() pliers of the onstraints P and, respetively Then KKT onditions are (b T t ) ([K t] kk k b tk η) + ψ k φ k =, k = 1,, M, where η = at t K t + b T t ψ k (k P )=, ψ k, k P, and φ k k =, ψ k, k, (19) Sine a T t K t, b t and >, we find η > Solving the KKT onditions yields k = { btk [K t] kk η, for η [Kt] kk b tk P, otherwise P () In Appendix E, we show that at least one of k s in () obtains its maximum P Suppose we sort the sensors suh b ti1 that [K t] i1 bt i M i 1 [K t] im, ie, i1 im Let i M = [i1,, im, im+1,, im ] T and i1 = = im = P, 1 m M Solving a T t K t + ηb T m t = for η, ombined with (), we find η = P j=1 [Kt]i j i j + m P j=1 bt i j If [Kt]imim b P tim η [Kt]i m+1 i m+1 b P tim+1, then the above assumption is valid, and we substitute η in () and alulate im+1,, im and MDC in (O 3 ) Note that η depends on m Otherwise, we inrease m by one and repeat the proedure, until we reah η that lies within the proper interval In Appendix E we also show that, although (O 3 ) is not onvex, the KKT solution in () is unique D Disussion on Maximization of MDC Using Convex Optimization Program Reall that in Setion IV-A we ould not find a losed form solution for (O 1 ) when all the entries of ˆq do not have the same sign Also, the analytial solution to (O ),

6 6 formulated in Setion IV-B, remains elusive Hene, we have provided a sub-optimal solution, via solving (O ) that is aurate solution when κ 1 Similarly, we have derived a sub-optimal solution to (O 3 ), formulated in Setion IV-C, that is aurate solution when ξ 1 In this setion, we turn (O 1 ),(O ),(O 3 ) into onvex optimization problems, in order to solve them numerially using CVX program We start with (O ), in whih we wish to minimize 1 = at t Ktat+ (b T t at), under TIPC Let x a = at b T t at MDC() t a = 1 b Therefore T bt t at t x a =1 and = xa and t a Employing these definitions, (O ) an be rewritten in the following equivalent form min x T a K t x a + t a (O 1 ) x a,t a st x T a x a P tot t a x a t a P x a b T t x a = 1 We an reformulate (O 1 ) as where min za T D a z a (O ) z a [ ] I st za T T z P a < 1 tot [I, P ]z a [b T t, ]z a = 1, z a z a = [x T a, t a ] T, D a = [ Kt T ] (1) Examining (O ), we realize that it is a quadrati programming (QP) onvex problem sine D a and hene it an be solved using CVX program One an take similar steps to turn (O 1 ) and (O 3 ) into a problem whose optimal solution an be found using CVX program In partiular, we formulate (O 11 ),(O 1 ) via deleting the seond inequality onstraints orresponding to IPC and (O 31 ),(O 3 ) by removing the first inequality onstraints orresponding to TPC from (O 1 ),(O ), respetively Sine D a, (O 1 ) and (O 3 ) are also QP onvex problems V NUMERICAL RESULTS In this setion, through simulations, we orroborate our analytial results We study the effet of orrelation between sensors observations on the MDC, the performane improvements ahieved by the MDC-based transmit power alloations (we refer to as DPA ), and the impat of different sensing and ommuniation hannels on DPA For our simulations, we onsider the signal model H : x k = z k, H 1 : x k = s k + z k, for k = 1,, M, where z k N (, σ ) and s k N (, σ s k ) is a sample of an external Gaussian signal soure s N (, σs) We assume σs k = σ s, where d d ɛs Sk is the distane between sensor k and s and ɛ s is the pathloss exponent S k We assume z k and s k are mutually unorrelated, however, s k s are orrelated Let s = [s 1, s,, s M ] T have ovariane matrix K s =E{ss T } We assume [K s ] ij =ρ ij σs i σs j where ρ ij = ρ dij, ρ 1 is the orrelation at unit distane and depends on the environment and d ij is the distane between sensors i and j [14] Eah sensor employs an energy detetor that maximizes p dk, under the onstraint p fk < 1 Sensors are deployed at equal distanes from eah other, on the irumferene of a irle with diameter m on the x-y plane, where the oordinate of its enter is (,, ) For sensing part, we assume M = 8, ɛ s =, σs = dbm, σ = 7 dbm, and for ommuniation part we let σn = 7 dbm, G = db [], ɛ =, and P 1 ==P M = P Performane of DPA when p dk s and pathloss are idential: Suppose the oordinates of signal soure s and the FC, respetively, are (,, 3m), and (,, m) With this onfiguration, p dk = 661, k and pathloss are idential We assume h k CN (, 1), k and we average over, number of hannel realizations to obtain the results We explore the MDC enhanements ahieved by DPA and ompare the MDC values with those of obtained by uniform power alloation (we refer to as UPA ), in whih sensors transmit at equal powers Fig 1 ompares optimal power alloation (OPA), whih finds the sensors powers that maximize P D under the onstraint P F < β F, DPA and UPA, for linear fusion rule and the optimal LRT rule To find OPA with both linear and the LRT rules and DPA with the LRT rule, we use brute fore searh to find the power values, and we simplify the network and only onsider s 1 and s We assume ρ = 1, β F = 1 and plot P D versus P tot under TPC for PAC Fig 1(a) ompares OPA, DPA and UPA, given linear fusion rule We observe that at low P tot, they are lose to eah other but as P tot inreases, they diverge and DPA outperforms UPA but performs worse than OPA Fig 1(b) ompares OPA, DPA and UPA, given the LRT rule Similarly, we see that DPA performs between UPA and OPA We note that at low P tot, DPA approahes OPA Also we plot P D for the LRT rule, where the transmit power values are obtained from maximizing the MDC of the linear fusion rule (in Fig 1(b) we refer to it as Power allo by DPA with linear rule ) We observe that, exept at low P tot, this urve is very lose to P D orresponding to DPA for the LRT rule, implying that the MDC-based power alloation with linear fusion rule is very lose to the MDC-based power alloation with LRT rule Figs and 3 show P D and maximized MDC versus P tot, respetively, under TPC for PAC and MAC, ρ = 1, 9, and β F = in Fig Comparing Figs and 3, we observe that the MDC and P D follow similar trends Hene, to make our omputations faster and less omplex, in the rest of this setion we only alulate the MDC From Fig 3, we note that the MDC inreases by inreasing P tot or by dereasing ρ Comparing PAC and MAC, we note that MAC outperforms PAC at low P tot, whereas PAC onverges to MAC at high P tot These are due to the fats that, at low P tot the effet of ommuniation hannel noise haraterized by in the MDC expression of PAC is M times larger than that of MAC (see equations () and (13)) and thus MAC outperforms PAC However, at high P tot this differene in values is negligible and hene PAC onverges to MAC We also observe that, at low P tot the performane gaps orresponding to DPA and UPA are negligible, despite the fat that sensors experiene different ommuniation hannel fading This is beause at low P tot the dominant effet of ommuniation hannel noise

7 7 renders the deisions of sensors equally important to the FC, regardless of the hannel realizations and the atual (different) deisions On the other hand, at high P tot the performane gaps orresponding to DPA and UPA are signifiant Note that this performane gap in MAC is wider than that of PAC This is expeted, sine the larger value in PAC undermines the differenes between sensors and narrows the performane gap between DPA and UPA As ρ inreases, the hanes that sensors make similar deisions inrease and therefore the performane gaps between DPA and UPA shrink Fig 4 shows the MDC maximized under TIPC versus P tot for PAC and MAC, P = 3mW, and ρ = 1, 9 Similar to Fig 3, the MDC inreases by inreasing P tot or dereasing ρ and MAC outperforms PAC We also ompare the MDC obtained from solving (O ) and (O ), in whih we have the inequality onstraint (I) P tot and the equality onstraint (E) = P tot, respetively We observe that at low P tot, there is no performane gap orresponding to DPA with E and DPA with I, whereas at high P tot, the performane of DPA with E degrades from that of DPA with I This performane degradation in MAC is due to the inreasing interferene of sensors deisions at the FC when sensors are assigned higher transmit power At very high P tot the performane of DPA with E redues to that of UPA This is beause the maximum value that P tot an assume is M P Hene, at very high P tot we have =M P, implying that k = P, k Fig shows the MDC maximized under IPC versus P for PAC and MAC and ρ=1, 9 We note that at low P, the performanes of DPA and UPA are similar, sine k = P, k This observation is in agreement with our analytial results in Setion IV-C, where we showed for ξ 1 we have = P Examining the effet of inreasing ρ from 1 to 9 in Figs 3, 4 and, we observe that the performane gap (ie, the differene between the two maximized MDC values) inreases as P tot or P inreases Trends of DPA when p dk s are different and pathloss are idential: We hange the oordinate of signal soure s to (m,, 3m), right above sensor S 1 With this onfiguration, p dk s hange to p T d = [739, 688, 683,, 37,, 683, 688], while pathloss are still idential (note that p d1, p d, p d8 are the three largest) Assuming h k = 1, k, we investigate the impat of different p dk s on DPA via plotting P tk, k Consider Fig 6 whih plots P tk for MAC Figs 6(a), 6(b) orrespond to the ase when the MDC is maximized under TPC, ρ = 1, 9, and P tot = 3 mw, 1 mw, 4 mw Figs 6(), 6(d) orrespond to the ase when the MDC is maximized under TIPC, ρ = 1, 9, P = 3 mw, and P tot = 3 mw, 1 mw, 4 mw Figs 6(e), 6(f) orrespond to the ase when the MDC is maximized under IPC, ρ = 1, 9, and P = 4 mw, 1 mw, 3 mw These figures show that for the ases when the MDC is maximized under TPC or under TIPC, sensors with higher p dk values (ie, more reliable loal deisions) are assigned higher P tk for all P tot values For the ase when the MDC is maximized under IPC, at low P, P tk = P, k (we have UPA) However, as P inreases, for those sensors with smaller p dk values (ie, less reliable loal deisions) we have smaller P tk We also note that as ρ inreases from 1 to 9 the variations of P tk aross sensors inrease: for the ase when the MDC is maximized under TPC, sensors with larger and smaller p dk s, respetively, are assigned further more and lesser P tk ; for the ase when the MDC is maximized under TIPC or IPC, sensors with smaller p dk s are assigned less P tk, suh that for p di < p dj we have P ti < P tj P Fig 7 plots P tk for PAC Comparing Figs 6 and 7, we note that similar trends hold true, while the variations of P tk s aross sensors in MAC, espeially in TIPC and IPC, are wider than those of PAC (ie, P tk s aross sensors in MAC are more different than UPA), due to the fat that the value in MAC is smaller Trends of DPA when p dk s are idential and pathloss are different: Suppose the oordinates of s and the FC, respetively, are (,, 3m), and (m,, 3m), where the FC is right below sensor S 1 With this onfiguration, p dk = 661, k, whereas the pathloss are different (note that the pathloss orresponding to S 1, S and S 8 are the three smallest) We observed that P tk s for different ρ values remain the same Hene, in this part we fous on ρ = 1 DPA is shown in figures 8 and 9 for MAC and PAC, respetively We observe that, for both PAC and MAC under TPC or TIPC, sensors with larger pathloss are assigned higher P tk (we refer to as inverse water filling) Examining the ase when the MDC is maximized under IPC and P = 4mW, we have P tk = P, k (UPA) in PAC, whereas sensors with larger pathloss are assigned higher P tk (inverse water filling) in MAC This is due to the fat that the value in MAC is smaller and therefore, the effetive reeived signal-to-noise ratio in MAC is larger, leading to variations of P tk s aross sensors To investigate more the effet of different pathloss on DPA, we move the FC further from the sensors and hange its oordinate to (m,, m), to effetively inrease the pathloss between all the sensors and the FC (and derease reeived power at the FC), while still S 1, S and S 8 have the three smallest pathloss We observe that in TPC and TIPC sensors with smaller pathloss are assigned higher P tk (we refer to as water filling), whereas in IPC, P tk = P, k (we have UPA) VI CONCLUSION We onsidered a hannel aware binary distributed detetion problem in a WSN with oherent reeption and linear fusion rule at the FC, where observations are orrelated Gaussian and sensors are unaware of suh orrelation when making deisions Assuming that the sensors and the FC are onneted via PAC or MAC, we studied power alloation shemes that maximize the MDC at the FC Our numerial results suggest that when MDC-based power alloation and optimal transmit power alloation are employed at low P tot, the resulting P D is very lose for both linear fusion rule and the LRT rule For homogeneous sensors with idential pathloss, MAC outperforms PAC at low P tot under TPC and TIPC (low P under IPC), whereas PAC onverges to MAC at high P tot Compared with equal power alloation, performane enhanement offered by the MDC-based power alloation is more signifiant in MAC and this improvement redues as orrelation inreases For inhomogeneous sensors with idential pathloss, sensors with more reliable deisions are assigned higher powers As

8 8 orrelation inreases, the variations of power aross sensors inrease: sensors with more (less) reliable deisions, are assigned further more (lesser) powers For homogeneous sensors with different pathloss, power alloations are invariant as orrelation hanges At low (high) reeived power at the FC, sensors with smaller (larger) pathloss are assigned higher powers under TPC and TIPC APPENDIX A Proof of Lemma Consider Q = DΛD T, where Λ = DIAG{[λ 1,, λ M ] T } and λ k s are the positive eigenvalues of Q and olumns of D are the eigenvetors of Q We an rewrite f(x) as f(x) = (x T (D Λ)(D Λ) 1 b t) x T D Λ = xt bt btt x ΛD T x x T x, where x = ΛD T x and b t =(D Λ) 1 b t Using the Rayleigh Ritz inequality [3], we find f(x) λ max ( b T t bt ) and the equality is ahieved when x is the orresponding eigenvetor of λ max ( b T t bt ) Sine T b t bt is rank-one with the eigenvalue bt and the eigenvetor b t, we have f(x) b tt bt = b T t Q 1 b t and the equality is ahieved at x = b t or x = Q 1 b t and its non-zero sales B Proving that solution of (O l ) satisfies TPC at the equality Consider (O) l and let µ and ψ, φ be the Lagrange multipliers orresponding to a T t P tot, P and, respetively The KKT onditions are b tk b T t a T t b t b T t µ k + ψ k φ k =, k = 1,, M () µ(a T t P tot ) =, µ, a T t P tot (3) ψ k (k P k ) =, ψ k, k P k, and φ k k =, φ k, k (4) We show µ Substituting µ = in (), we have b tk b T t at a T t btbt t at = φ k ψ k Sine b t,,, we find bt k bt t at = φ a T k ψ k < Note that ψ k and φ k t btbt t annot be both at positive, sine from (4) it is infeasible to have k = P k and k = Therefore ψ k or φ k must be zero Sine φ k ψ k <, we onlude that φ k = and ψ k > Now, from (4) we have k = P k, leading to a T t = M k=1 P k > P tot, whih ontradits (3) Therefore, µ and we have a T t =P tot C Analytial Solution of (O ) Sine a T t1 a t1 =P tot, the objetive funtion in (O ) redues to P tot a T t1 Let µ and ψ and φ be the Lagrange multipliers orresponding to a T t = P tot, P and The KKT onditions are µk a t1 k + ψ k φ k =, k = 1,, M µ(a T t P tot ) =, a T t = P tot ψ k (k P k ) =, ψ k, k P k, and φ k k =, φ k, k Solving the KKT onditions for a t1 yields a t1 k µ, for µ a t1 k P k k = Pk, for µ < a t1 k P k () To find positive µ and onsequently k, suppose sensors are a t1 sorted suh that i1 Pi1 a t1 im PiM Assume for 1 m M we have i1 = P i1,, im = P im Substituting () into M j=1 ij = P tot and solving for µ, we find µ = M j=m+1 a t1 ij 4(P tot m j=1 P i j ) Note that µ depends on m If a t1 im+1 P im+1 µ a t1 im, the above assumption is valid, and we substitute P im µ in () to alulate im+1,, im Otherwise, we inrease m by one and repeat the proedure, until we reah µ that lies within the proper interval Although (O 3 ) is not onvex, we show below that the KKT solution in () is unique Suppose the solution in () is not unique, ie, there exist 1 m, m M, m m + 1 suh that ij = a t1 ij µ at1 ij, for m + 1 j M, and µ, P ij Pij, for 1 j m, and µ < a t1 ij P ij (6) Also, ij an be obtained from (6) by substituting j, m, µ with j, m, µ, respetively Sine m m + 1, from (6), we have µ a t1 im+1 4P im+1 On the other hand, due to sensor ordering, we have a t1 im+1 4P im+1 a inequality, we obtain t1 im 4P im m j=m+1 a t1 ij 4 m j=m+1 P i j Applying the mediant a t1 im+1 4P im+1 We observe that two different frations are greater than or equal to the a t1 im+1 4P im+1 Hene, using the mediant inequality and definition of µ, we find M j=m+1 a t1 ij m j=m+1 a t1 ij 4(P tot m j=1 P ij m j=m+1 P ij ) = µ a t1 im+1 4P im+1 However, this inequality ontradits the one in (6) when j, µ are replaed with j, µ Hene, our assumption regarding the existene of m, m is inorret and the solution in () is unique D Proving that solution of (O l 3) is equal to IPC upper limit Consider (O l 3) and let ψ, φ be the Lagrange multipliers orresponding to P and, respetively The KKT onditions are b tk b T t a T t b t b T t + ψ k φ k =, k = 1,, M ψ k (k P k ) =, ψ k, k P k, and φ k k =, φ k, k (7) Sine b t,,, we find bt k bt t at a T t btbt t at =φ k ψ k < Note that ψ k and φ k annot be both positive, sine from (7) it is infeasible to have k = P k and k = Therefore either ψ k or φ k must be zero Sine φ k ψ k <, we onlude that φ k = and ψ k > Now, from (7) we have k = P k, or equivalently, = P

9 9 E Regarding Analytial Solution to (O 3 ) with Independent Observations First, we show that at least one of k s in () is equal to P Suppose k < P, k From (), we have k = bt k [K t] kk η or equivalently η = at k [Kt] kkk, k Using the b tk k M k=1 mediant inequality, we rewrite η as η = at k [Kt] kkk M = k=1 bt k at k a T t Ktat < at b T t Ktat+, sine > However, this violates the t at b T t definition of η in at (19) and ontradits our initial assumption Hene, there should be at least one k = P Next, we show that, although (O 3 ) is not onvex, the KKT solution in () is unique Suppose sensors are sorted suh that b ti1 [K t] i1 bt i M i 1 [K t] im, ie, i1 im, where at least i M i1 = P Assume that the solution in () is not unique, ie, there exist two indies m, m {1,, M}, m m + 1 suh that ij = b tij [K t] ij η, for m + 1 j M, η [Kt]i j i j i j b tij P, for 1 j m, and η > [Kt]i j i j b tij P P where η = P m l=1 [Kt]i l i l + P m l=1 bt i l Also, ij an be obtained from above by substituting j, m, η with j, m, η, respetively Sine m m + 1, from (8), we find η [Kt]i m i m b P tim On the other hand, due to sensor ordering, we have bt i m [K t] im i m P m l=m+1 [Kt]i l i l P m l=m+1 bt i l b tim+1 [K t] im+1 i m+1 Applying the mediant inequality, we obtain P We observe that two [K t] i m i m b tim different frations are less than or equal to [Kt]i m i m b tim Thus m P l=1 [K m t] il i l + + P l=m+1 [K t] il i l m P l=1 b t + m =η il P [K t ] im i m b tim P l=m+1 b t il P However, this inequality ontradits the one in (8) when j, η are replaed with j, η Hene, our assumption regarding the existene of two indies m, m is inorret and the solution in () is unique REFERENCES [1] P K Varshney, Distributed Detetion and Data Fusion New York: Springer, 1997 [] R Viswanathan and P K Varshney, Distributed detetion with multiple sensors: part I-fundamentals, Proeedings of the IEEE, vol 8, pp 64 79, Jan 1997 [3] Q Yan and R S Blum, Distributed signal detetion under the neymanpearson riterion, IEEE Trans Information Theory, vol 47, no 4, pp , 1 [4] B Chen, L Tong, and P Varshney, Channel-aware distributed detetion in wireless sensor networks, IEEE Signal Proessing Mag, vol 3, pp 16 6, July 6 [] B Chen, R Jiang, T Kasetkasem, and P Varshney, Channel aware deision fusion in wireless sensor networks, IEEE Trans Signal Proessing, vol, pp , De 4 [6] N Ruixin, B Chen, and P K Varshney, Fusion of deisions transmitted over Rayleigh fading hannels in wireless sensor networks, IEEE Trans Signal Proessing, vol 4, pp 18 7, Mar 6 [7] H R Ahmadi and A Vosoughi, Optimal training and data power alloation in distributed detetion with inhomogeneous sensors, IEEE Signal Proessing Letters, vol, no 4, pp , 13 [8], Distributed detetion with adaptive topology and nonideal ommuniation hannels, IEEE Trans Signal Proessing, vol 9, pp , June 11 [9], Impat of wireless hannel unertainty upon distributed detetion systems, IEEE Trans Wireless Comm, vol 1, pp 66 77, June 13 [] X Zhang, H Poor, and M Chiang, Optimal power alloation for distributed detetion over mimo hannels in wireless sensor networks, IEEE Trans Signal Proessing, vol 6, no 9, pp , Sept 8 [11] h S Kim, J Wang, P Cai, and S Cui, Detetion outage and detetion diversity in a homogeneous distributed sensor network, IEEE Trans Signal Proessing, vol 7, no 7, pp , July 9 [1] C R Berger, M Guerriero, S Zhou, and P Willett, PAC vs MAC for deentralized detetion using nonoherent modulation, IEEE Trans Signal Proessing, vol 7, no 9, pp 36 37, Sept 9 [13] J J Xiao, S Cui, Z Q Luo, and A J Goldsmith, Linear oherent deentralized estimation, IEEE Trans Signal Proessing, vol 6, no, pp 77 77, Feb 8 [14] M H Chaudhary and L Vandendorpe, Power onstrained linear estimation in wireless sensor networks wirh orrelated data and digital modulation, IEEE Trans Signal Proessing, vol 6, no, pp 7 84, Feb 1 [1] S Cui, J Xiao, A J Goldsmith, Z Q Luo, and H V Poor, Estimation diversity and energy effiieny in distributed sensing, IEEE Trans Signal Proessing, vol, no 9, pp , Sep 7 [16] J Fang and H Li, Power onstrained distributed estimation with orrelated sensor data, IEEE Trans Signal Proessing, vol 7, no 8, pp , Aug 9 [17] J Y Wu and T Y Wang, Power alloation for robust distributed bestlinear-unbiased estimation against sensing noise variane unertainty, IEEE Trans Wireless Communiations, vol 1, no 6, pp , 13 [18] G Alirezaei, M Reyer, and R Mathar, Optimum power alloation in sensor networks for ative radar appliations, IEEE Trans Wireless Communiations, vol 14, no, pp , 1 [19] S Kar and P K Varshney, Linear oherent estimation with spatial ollaboration, IEEE Trans Info Theory, vol 9, pp 33 33, June 13 [] S Liu, S Kar, M Fardad, and P K Varshney, Optimized sensor ollaboration for estimation of temporally orrelated parameters, IEEE Trans Signal Proessing, vol 64, pp , De 16 [1] A S Behbahani, A M Eltawil, and H Jafarkhani, Linear deentralized estimation of orrelated data for power-onstrained wireless sensor networks, IEEE Trans Signal Proessing, vol 6, no 11, pp , 1 [] I Bahei and A K Khandani, Linear estimation of orrelated data in wireless sensor networks with optimum power alloation and analog modulation, IEEE Trans Communiations,, vol 6, no 7, pp , 8 [3] P Willet, P Swaszek, and R S Blum, The good, bad and ugly: distributed detetion of a known signal in dependent Gaussian noise, IEEE Trans Signal Proessing, vol 48, pp , De [4] H Chen, B Chen, and P K Varshney, A new framework for distributed detetion with onditionally dependent observations, IEEE Trans Signal Proessing, vol 6, no 3, pp , 1 [] Y Lin and H Chen, Distributed detetion performane under dependent observations and nonideal hannels, Sensors Journal, IEEE, vol 1, no, pp 71 7, 1 [6] Z Quan, S Cui, and A H Sayed, Optimal linear ooperation for spetrum sensing in ognitive radio networks, IEEE J Sel Topis in Signal Proessing, vol, pp 8 4, Feb 8 [7] B Piinbono, On defletion as a performane riterion in detetion, IEEE Trans Aerospae and Eletroni Systems, vol 31, no 3, pp 7 81, July 199 [8] J Unnikrishnan and V V Veeravalli, Cooperative sensing for primary detetion in ognitive radio, IEEE J Sel Topis in Signal Proessing, vol, no 1, pp 18 7, February 8 [9] K C Lai, Y L Yang, and J J Jia, Fusion of deisions transmitted over flat fading hannels via maximizing the defletion oeffiient, IEEE Trans Vehiular Tehnology, vol 9, pp , Sept [3] J R Magnus and H Neudeker, Matrix Differential Calulus with Appliations in Statistis and Eonometris Wiley; nd edition, 1999

10 OPA DPA UPA 6 P D Total transmit power (mw) (a) P D OPA Power allo by DPA with linear rule DPA UPA Total transmit power (mw) Fig 1 P D under TPC versus P tot for a -sensor PAC with idential p dk s and pathloss and ρ=1: (a) Linear fusion rule, (b) LRT fusion rule (b)

11 P D PAC DPA =1 PAC UPA =1 PAC DPA =9 PAC UPA =9 MAC DPA =1 MAC UPA =1 MAC DPA =9 MAC UPA =9 1 Total transmit power (mw) Fig P D under TPC versus P tot 1 8 MDC 6 4 PAC DPA =1 PAC UPA =1 PAC DPA =9 PAC UPA =9 MAC DPA =1 MAC UPA =1 MAC DPA =9 MAC UPA =9 1 Total transmit power (mw) Fig 3 Maximized MDC under TPC versus P tot

12 1 9 MDC PAC DPA I =1 PAC UPA =1 PAC DPA E =1 PAC DPA I =9 PAC UPA =9 PAC DPA E =1 MAC DPA I =1 MAC UPA =1 MAC DPA E =1 MAC DPA I =9 MAC UPA =9 MAC DPA E =9 3 1 Total transmit power (mw) Fig 4 Maximized MDC under TIPC versus P tot and P = 3 mw MDC PAC DPA =1 PAC UPA =1 PAC DPA =9 PAC UPA =9 MAC DPA =1 MAC UPA =1 MAC DPA =9 MAC UPA =9 1 3 Individual transmit power (mw) Fig Maximized MDC under IPC versus P

13 Total power: 3mW Total power: 1mW Total power: 4mW Total power: 3mW Total power: 1mW Total power: 4mW (a) (b) Total power: 3mW Total power: 1mW Total power: 4mW 1 Total power: 3mW Total power: 1mW Total power: 4mW () (d) 3 3 Max Individ power: 4mW Max Individ power: 4mW 1 Max Individ power: 1mW Max Individ power: 3mW 1 Max Individ power: 1mW Max Individ power: 3mW (e) (f) Fig 6 DPA in MAC with different p dk s and idential pathloss: (a) Maximized MDC under TPC, ρ = 1; (b) Maximized MDC under TPC, ρ = 9; () Maximized MDC under TIPC, ρ = 1, P = 3 mw, (d) Maximized MDC under TIPC, ρ = 9, P = 3 mw; (e) Maximized MDC under IPC, ρ = 1, (f) Maximized MDC under IPC, ρ = 9

14 Total power: 3mW Total power: 1mW Total power: 4mW Total power: 3mW Total power: 1mW Total power: 4mW (a) (b) Total power: 3mW Total power: 1mW Total power: 4mW 1 Total power: 3mW Total power: 1mW Total power: 4mW () (d) 3 3 Max Individ power: 4mW 1 Max Individ power: 4mW Max Individ power: 1mW Max Individ power: 3mW 1 Max Individ power: 1mW Max Individ power: 3mW (e) (f) Fig 7 DPA in PAC with different p dk s and idential pathloss: (a) Maximized MDC under TPC, ρ = 1; (b) Maximized MDC under TPC, ρ = 9; () Maximized MDC under TIPC, ρ = 1, P = 3 mw, (d) Maximized MDC under TIPC, ρ = 9, P = 3 mw; (e) Maximized MDC under IPC, ρ = 1, (f) Maximized MDC under IPC, ρ = 9

15 Total power: 3mW Total power: 1mW Total power: 4mW 1 Total power: 3mW Total power: 1mW Total power: 4mW (a) (b) 3 Max Individ power: 4mW 1 Max Individ power: 1mW Max Individ power: 3mW () Fig 8 DPA in MAC with idential p dk s, different pathloss and ρ=1: (a) Maximized MDC under TPC, (b) Maximized MDC under TIPC, P =3 mw, () Maximized MDC under IPC Total power: 3mW Total power: 1mW Total power: 4mW 1 Total power: 3mW Total power: 1mW Total power: 4mW (a) (b) 3 1 Max Individ power: 4mW Max Individ power: 1mW Max Individ power: 3mW () Fig 9 DPA in PAC with idential p dk s, different pathloss and ρ=1: (a) Maximized MDC under TPC, (b) Maximized MDC under TIPC, P =3 mw, () Maximized MDC under IPC

Case I: 2 users In case of 2 users, the probability of error for user 1 was earlier derived to be 2 A1

Case I: 2 users In case of 2 users, the probability of error for user 1 was earlier derived to be 2 A1 MUTLIUSER DETECTION (Letures 9 and 0) 6:33:546 Wireless Communiations Tehnologies Instrutor: Dr. Narayan Mandayam Summary By Shweta Shrivastava (shwetash@winlab.rutgers.edu) bstrat This artile ontinues

More information

Advanced Computational Fluid Dynamics AA215A Lecture 4

Advanced Computational Fluid Dynamics AA215A Lecture 4 Advaned Computational Fluid Dynamis AA5A Leture 4 Antony Jameson Winter Quarter,, Stanford, CA Abstrat Leture 4 overs analysis of the equations of gas dynamis Contents Analysis of the equations of gas

More information

A Spatiotemporal Approach to Passive Sound Source Localization

A Spatiotemporal Approach to Passive Sound Source Localization A Spatiotemporal Approah Passive Sound Soure Loalization Pasi Pertilä, Mikko Parviainen, Teemu Korhonen and Ari Visa Institute of Signal Proessing Tampere University of Tehnology, P.O.Box 553, FIN-330,

More information

Assessing the Performance of a BCI: A Task-Oriented Approach

Assessing the Performance of a BCI: A Task-Oriented Approach Assessing the Performane of a BCI: A Task-Oriented Approah B. Dal Seno, L. Mainardi 2, M. Matteui Department of Eletronis and Information, IIT-Unit, Politenio di Milano, Italy 2 Department of Bioengineering,

More information

10.5 Unsupervised Bayesian Learning

10.5 Unsupervised Bayesian Learning The Bayes Classifier Maximum-likelihood methods: Li Yu Hongda Mao Joan Wang parameter vetor is a fixed but unknown value Bayes methods: parameter vetor is a random variable with known prior distribution

More information

Capacity-achieving Input Covariance for Correlated Multi-Antenna Channels

Capacity-achieving Input Covariance for Correlated Multi-Antenna Channels Capaity-ahieving Input Covariane for Correlated Multi-Antenna Channels Antonia M. Tulino Universita Degli Studi di Napoli Federio II 85 Napoli, Italy atulino@ee.prineton.edu Angel Lozano Bell Labs (Luent

More information

Methods of evaluating tests

Methods of evaluating tests Methods of evaluating tests Let X,, 1 Xn be i.i.d. Bernoulli( p ). Then 5 j= 1 j ( 5, ) T = X Binomial p. We test 1 H : p vs. 1 1 H : p>. We saw that a LRT is 1 if t k* φ ( x ) =. otherwise (t is the observed

More information

Complexity of Regularization RBF Networks

Complexity of Regularization RBF Networks Complexity of Regularization RBF Networks Mark A Kon Department of Mathematis and Statistis Boston University Boston, MA 02215 mkon@buedu Leszek Plaskota Institute of Applied Mathematis University of Warsaw

More information

Millennium Relativity Acceleration Composition. The Relativistic Relationship between Acceleration and Uniform Motion

Millennium Relativity Acceleration Composition. The Relativistic Relationship between Acceleration and Uniform Motion Millennium Relativity Aeleration Composition he Relativisti Relationship between Aeleration and niform Motion Copyright 003 Joseph A. Rybzyk Abstrat he relativisti priniples developed throughout the six

More information

Chapter Review of of Random Processes

Chapter Review of of Random Processes Chapter.. Review of of Random Proesses Random Variables and Error Funtions Conepts of Random Proesses 3 Wide-sense Stationary Proesses and Transmission over LTI 4 White Gaussian Noise Proesses @G.Gong

More information

Danielle Maddix AA238 Final Project December 9, 2016

Danielle Maddix AA238 Final Project December 9, 2016 Struture and Parameter Learning in Bayesian Networks with Appliations to Prediting Breast Caner Tumor Malignany in a Lower Dimension Feature Spae Danielle Maddix AA238 Final Projet Deember 9, 2016 Abstrat

More information

Sensitivity of Spectrum Sensing Techniques to RF impairments

Sensitivity of Spectrum Sensing Techniques to RF impairments Sensitivity of Spetrum Sensing Tehniques to RF impairments Jonathan Verlant-Chenet Julien Renard Jean-Mihel Driot Philippe De Donker François Horlin Université Libre de Bruelles - OPERA Dpt., Avenue F.D.

More information

Hankel Optimal Model Order Reduction 1

Hankel Optimal Model Order Reduction 1 Massahusetts Institute of Tehnology Department of Eletrial Engineering and Computer Siene 6.245: MULTIVARIABLE CONTROL SYSTEMS by A. Megretski Hankel Optimal Model Order Redution 1 This leture overs both

More information

Model-based mixture discriminant analysis an experimental study

Model-based mixture discriminant analysis an experimental study Model-based mixture disriminant analysis an experimental study Zohar Halbe and Mayer Aladjem Department of Eletrial and Computer Engineering, Ben-Gurion University of the Negev P.O.Box 653, Beer-Sheva,

More information

On the Bit Error Probability of Noisy Channel Networks With Intermediate Node Encoding I. INTRODUCTION

On the Bit Error Probability of Noisy Channel Networks With Intermediate Node Encoding I. INTRODUCTION 5188 IEEE TRANSACTIONS ON INFORMATION THEORY, VOL. 54, NO. 11, NOVEMBER 2008 [8] A. P. Dempster, N. M. Laird, and D. B. Rubin, Maximum likelihood estimation from inomplete data via the EM algorithm, J.

More information

A model for measurement of the states in a coupled-dot qubit

A model for measurement of the states in a coupled-dot qubit A model for measurement of the states in a oupled-dot qubit H B Sun and H M Wiseman Centre for Quantum Computer Tehnology Centre for Quantum Dynamis Griffith University Brisbane 4 QLD Australia E-mail:

More information

The Effectiveness of the Linear Hull Effect

The Effectiveness of the Linear Hull Effect The Effetiveness of the Linear Hull Effet S. Murphy Tehnial Report RHUL MA 009 9 6 Otober 009 Department of Mathematis Royal Holloway, University of London Egham, Surrey TW0 0EX, England http://www.rhul.a.uk/mathematis/tehreports

More information

Wave Propagation through Random Media

Wave Propagation through Random Media Chapter 3. Wave Propagation through Random Media 3. Charateristis of Wave Behavior Sound propagation through random media is the entral part of this investigation. This hapter presents a frame of referene

More information

Measuring & Inducing Neural Activity Using Extracellular Fields I: Inverse systems approach

Measuring & Inducing Neural Activity Using Extracellular Fields I: Inverse systems approach Measuring & Induing Neural Ativity Using Extraellular Fields I: Inverse systems approah Keith Dillon Department of Eletrial and Computer Engineering University of California San Diego 9500 Gilman Dr. La

More information

A NETWORK SIMPLEX ALGORITHM FOR THE MINIMUM COST-BENEFIT NETWORK FLOW PROBLEM

A NETWORK SIMPLEX ALGORITHM FOR THE MINIMUM COST-BENEFIT NETWORK FLOW PROBLEM NETWORK SIMPLEX LGORITHM FOR THE MINIMUM COST-BENEFIT NETWORK FLOW PROBLEM Cen Çalışan, Utah Valley University, 800 W. University Parway, Orem, UT 84058, 801-863-6487, en.alisan@uvu.edu BSTRCT The minimum

More information

Product Policy in Markets with Word-of-Mouth Communication. Technical Appendix

Product Policy in Markets with Word-of-Mouth Communication. Technical Appendix rodut oliy in Markets with Word-of-Mouth Communiation Tehnial Appendix August 05 Miro-Model for Inreasing Awareness In the paper, we make the assumption that awareness is inreasing in ustomer type. I.e.,

More information

Phase Diffuser at the Transmitter for Lasercom Link: Effect of Partially Coherent Beam on the Bit-Error Rate.

Phase Diffuser at the Transmitter for Lasercom Link: Effect of Partially Coherent Beam on the Bit-Error Rate. Phase Diffuser at the Transmitter for Laserom Link: Effet of Partially Coherent Beam on the Bit-Error Rate. O. Korotkova* a, L. C. Andrews** a, R. L. Phillips*** b a Dept. of Mathematis, Univ. of Central

More information

Maximum Entropy and Exponential Families

Maximum Entropy and Exponential Families Maximum Entropy and Exponential Families April 9, 209 Abstrat The goal of this note is to derive the exponential form of probability distribution from more basi onsiderations, in partiular Entropy. It

More information

Lecture 7: Sampling/Projections for Least-squares Approximation, Cont. 7 Sampling/Projections for Least-squares Approximation, Cont.

Lecture 7: Sampling/Projections for Least-squares Approximation, Cont. 7 Sampling/Projections for Least-squares Approximation, Cont. Stat60/CS94: Randomized Algorithms for Matries and Data Leture 7-09/5/013 Leture 7: Sampling/Projetions for Least-squares Approximation, Cont. Leturer: Mihael Mahoney Sribe: Mihael Mahoney Warning: these

More information

Taste for variety and optimum product diversity in an open economy

Taste for variety and optimum product diversity in an open economy Taste for variety and optimum produt diversity in an open eonomy Javier Coto-Martínez City University Paul Levine University of Surrey Otober 0, 005 María D.C. Garía-Alonso University of Kent Abstrat We

More information

Remark 4.1 Unlike Lyapunov theorems, LaSalle s theorem does not require the function V ( x ) to be positive definite.

Remark 4.1 Unlike Lyapunov theorems, LaSalle s theorem does not require the function V ( x ) to be positive definite. Leture Remark 4.1 Unlike Lyapunov theorems, LaSalle s theorem does not require the funtion V ( x ) to be positive definite. ost often, our interest will be to show that x( t) as t. For that we will need

More information

Nonreversibility of Multiple Unicast Networks

Nonreversibility of Multiple Unicast Networks Nonreversibility of Multiple Uniast Networks Randall Dougherty and Kenneth Zeger September 27, 2005 Abstrat We prove that for any finite direted ayli network, there exists a orresponding multiple uniast

More information

1 sin 2 r = 1 n 2 sin 2 i

1 sin 2 r = 1 n 2 sin 2 i Physis 505 Fall 005 Homework Assignment #11 Solutions Textbook problems: Ch. 7: 7.3, 7.5, 7.8, 7.16 7.3 Two plane semi-infinite slabs of the same uniform, isotropi, nonpermeable, lossless dieletri with

More information

Word of Mass: The Relationship between Mass Media and Word-of-Mouth

Word of Mass: The Relationship between Mass Media and Word-of-Mouth Word of Mass: The Relationship between Mass Media and Word-of-Mouth Roman Chuhay Preliminary version Marh 6, 015 Abstrat This paper studies the optimal priing and advertising strategies of a firm in the

More information

Sensitivity Analysis in Markov Networks

Sensitivity Analysis in Markov Networks Sensitivity Analysis in Markov Networks Hei Chan and Adnan Darwihe Computer Siene Department University of California, Los Angeles Los Angeles, CA 90095 {hei,darwihe}@s.ula.edu Abstrat This paper explores

More information

Exploring the feasibility of on-site earthquake early warning using close-in records of the 2007 Noto Hanto earthquake

Exploring the feasibility of on-site earthquake early warning using close-in records of the 2007 Noto Hanto earthquake Exploring the feasibility of on-site earthquake early warning using lose-in reords of the 2007 Noto Hanto earthquake Yih-Min Wu 1 and Hiroo Kanamori 2 1. Department of Geosienes, National Taiwan University,

More information

The Hanging Chain. John McCuan. January 19, 2006

The Hanging Chain. John McCuan. January 19, 2006 The Hanging Chain John MCuan January 19, 2006 1 Introdution We onsider a hain of length L attahed to two points (a, u a and (b, u b in the plane. It is assumed that the hain hangs in the plane under a

More information

Probabilistic Graphical Models

Probabilistic Graphical Models Probabilisti Graphial Models David Sontag New York University Leture 12, April 19, 2012 Aknowledgement: Partially based on slides by Eri Xing at CMU and Andrew MCallum at UMass Amherst David Sontag (NYU)

More information

DIGITAL DISTANCE RELAYING SCHEME FOR PARALLEL TRANSMISSION LINES DURING INTER-CIRCUIT FAULTS

DIGITAL DISTANCE RELAYING SCHEME FOR PARALLEL TRANSMISSION LINES DURING INTER-CIRCUIT FAULTS CHAPTER 4 DIGITAL DISTANCE RELAYING SCHEME FOR PARALLEL TRANSMISSION LINES DURING INTER-CIRCUIT FAULTS 4.1 INTRODUCTION Around the world, environmental and ost onsiousness are foring utilities to install

More information

QCLAS Sensor for Purity Monitoring in Medical Gas Supply Lines

QCLAS Sensor for Purity Monitoring in Medical Gas Supply Lines DOI.56/sensoren6/P3. QLAS Sensor for Purity Monitoring in Medial Gas Supply Lines Henrik Zimmermann, Mathias Wiese, Alessandro Ragnoni neoplas ontrol GmbH, Walther-Rathenau-Str. 49a, 7489 Greifswald, Germany

More information

A Unified View on Multi-class Support Vector Classification Supplement

A Unified View on Multi-class Support Vector Classification Supplement Journal of Mahine Learning Researh??) Submitted 7/15; Published?/?? A Unified View on Multi-lass Support Vetor Classifiation Supplement Ürün Doğan Mirosoft Researh Tobias Glasmahers Institut für Neuroinformatik

More information

Research Collection. Mismatched decoding for the relay channel. Conference Paper. ETH Library. Author(s): Hucher, Charlotte; Sadeghi, Parastoo

Research Collection. Mismatched decoding for the relay channel. Conference Paper. ETH Library. Author(s): Hucher, Charlotte; Sadeghi, Parastoo Researh Colletion Conferene Paper Mismathed deoding for the relay hannel Author(s): Huher, Charlotte; Sadeghi, Parastoo Publiation Date: 2010 Permanent Link: https://doi.org/10.3929/ethz-a-005997152 Rights

More information

Analysis of discretization in the direct simulation Monte Carlo

Analysis of discretization in the direct simulation Monte Carlo PHYSICS OF FLUIDS VOLUME 1, UMBER 1 OCTOBER Analysis of disretization in the diret simulation Monte Carlo iolas G. Hadjionstantinou a) Department of Mehanial Engineering, Massahusetts Institute of Tehnology,

More information

Singular Event Detection

Singular Event Detection Singular Event Detetion Rafael S. Garía Eletrial Engineering University of Puerto Rio at Mayagüez Rafael.Garia@ee.uprm.edu Faulty Mentor: S. Shankar Sastry Researh Supervisor: Jonathan Sprinkle Graduate

More information

Enhanced Max-Min SINR for Uplink Cell-Free Massive MIMO Systems

Enhanced Max-Min SINR for Uplink Cell-Free Massive MIMO Systems Enhaned Max-Min SINR for Uplin Cell-Free Massive MIMO Systems Manijeh Bashar, Kanapathippillai Cumanan, Alister G. Burr,Mérouane Debbah, and Hien Quo Ngo Department of Eletroni Engineering, University

More information

Connectivity and Blockage Effects in Millimeter-Wave Air-To-Everything Networks

Connectivity and Blockage Effects in Millimeter-Wave Air-To-Everything Networks 1 Connetivity and Blokage Effets in Millimeter-Wave Air-To-Everything Networks Kaifeng Han, Kaibin Huang and Robert W. Heath Jr. arxiv:1808.00144v1 [s.it] 1 Aug 2018 Abstrat Millimeter-wave (mmwave) offers

More information

Reliability Guaranteed Energy-Aware Frame-Based Task Set Execution Strategy for Hard Real-Time Systems

Reliability Guaranteed Energy-Aware Frame-Based Task Set Execution Strategy for Hard Real-Time Systems Reliability Guaranteed Energy-Aware Frame-Based ask Set Exeution Strategy for Hard Real-ime Systems Zheng Li a, Li Wang a, Shuhui Li a, Shangping Ren a, Gang Quan b a Illinois Institute of ehnology, Chiago,

More information

the following action R of T on T n+1 : for each θ T, R θ : T n+1 T n+1 is defined by stated, we assume that all the curves in this paper are defined

the following action R of T on T n+1 : for each θ T, R θ : T n+1 T n+1 is defined by stated, we assume that all the curves in this paper are defined How should a snake turn on ie: A ase study of the asymptoti isoholonomi problem Jianghai Hu, Slobodan N. Simić, and Shankar Sastry Department of Eletrial Engineering and Computer Sienes University of California

More information

Subject: Introduction to Component Matching and Off-Design Operation % % ( (1) R T % (

Subject: Introduction to Component Matching and Off-Design Operation % % ( (1) R T % ( 16.50 Leture 0 Subjet: Introdution to Component Mathing and Off-Design Operation At this point it is well to reflet on whih of the many parameters we have introdued (like M, τ, τ t, ϑ t, f, et.) are free

More information

Non-Markovian study of the relativistic magnetic-dipole spontaneous emission process of hydrogen-like atoms

Non-Markovian study of the relativistic magnetic-dipole spontaneous emission process of hydrogen-like atoms NSTTUTE OF PHYSCS PUBLSHNG JOURNAL OF PHYSCS B: ATOMC, MOLECULAR AND OPTCAL PHYSCS J. Phys. B: At. Mol. Opt. Phys. 39 ) 7 85 doi:.88/953-75/39/8/ Non-Markovian study of the relativisti magneti-dipole spontaneous

More information

arxiv:gr-qc/ v2 6 Feb 2004

arxiv:gr-qc/ v2 6 Feb 2004 Hubble Red Shift and the Anomalous Aeleration of Pioneer 0 and arxiv:gr-q/0402024v2 6 Feb 2004 Kostadin Trenčevski Faulty of Natural Sienes and Mathematis, P.O.Box 62, 000 Skopje, Maedonia Abstrat It this

More information

Where as discussed previously we interpret solutions to this partial differential equation in the weak sense: b

Where as discussed previously we interpret solutions to this partial differential equation in the weak sense: b Consider the pure initial value problem for a homogeneous system of onservation laws with no soure terms in one spae dimension: Where as disussed previously we interpret solutions to this partial differential

More information

A Queueing Model for Call Blending in Call Centers

A Queueing Model for Call Blending in Call Centers A Queueing Model for Call Blending in Call Centers Sandjai Bhulai and Ger Koole Vrije Universiteit Amsterdam Faulty of Sienes De Boelelaan 1081a 1081 HV Amsterdam The Netherlands E-mail: {sbhulai, koole}@s.vu.nl

More information

An Adaptive Optimization Approach to Active Cancellation of Repeated Transient Vibration Disturbances

An Adaptive Optimization Approach to Active Cancellation of Repeated Transient Vibration Disturbances An aptive Optimization Approah to Ative Canellation of Repeated Transient Vibration Disturbanes David L. Bowen RH Lyon Corp / Aenteh, 33 Moulton St., Cambridge, MA 138, U.S.A., owen@lyonorp.om J. Gregory

More information

Optimization of Statistical Decisions for Age Replacement Problems via a New Pivotal Quantity Averaging Approach

Optimization of Statistical Decisions for Age Replacement Problems via a New Pivotal Quantity Averaging Approach Amerian Journal of heoretial and Applied tatistis 6; 5(-): -8 Published online January 7, 6 (http://www.sienepublishinggroup.om/j/ajtas) doi:.648/j.ajtas.s.65.4 IN: 36-8999 (Print); IN: 36-96 (Online)

More information

Lecture 3 - Lorentz Transformations

Lecture 3 - Lorentz Transformations Leture - Lorentz Transformations A Puzzle... Example A ruler is positioned perpendiular to a wall. A stik of length L flies by at speed v. It travels in front of the ruler, so that it obsures part of the

More information

SURFACE WAVES OF NON-RAYLEIGH TYPE

SURFACE WAVES OF NON-RAYLEIGH TYPE SURFACE WAVES OF NON-RAYLEIGH TYPE by SERGEY V. KUZNETSOV Institute for Problems in Mehanis Prosp. Vernadskogo, 0, Mosow, 75 Russia e-mail: sv@kuznetsov.msk.ru Abstrat. Existene of surfae waves of non-rayleigh

More information

Wavetech, LLC. Ultrafast Pulses and GVD. John O Hara Created: Dec. 6, 2013

Wavetech, LLC. Ultrafast Pulses and GVD. John O Hara Created: Dec. 6, 2013 Ultrafast Pulses and GVD John O Hara Created: De. 6, 3 Introdution This doument overs the basi onepts of group veloity dispersion (GVD) and ultrafast pulse propagation in an optial fiber. Neessarily, it

More information

Research Letter Distributed Source Localization Based on TOA Measurements in Wireless Sensor Networks

Research Letter Distributed Source Localization Based on TOA Measurements in Wireless Sensor Networks Researh Letters in Eletronis Volume 2009, Artile ID 573129, 4 pages doi:10.1155/2009/573129 Researh Letter Distributed Soure Loalization Based on TOA Measurements in Wireless Sensor Networks Wanzhi Qiu

More information

Stochastic Combinatorial Optimization with Risk Evdokia Nikolova

Stochastic Combinatorial Optimization with Risk Evdokia Nikolova Computer Siene and Artifiial Intelligene Laboratory Tehnial Report MIT-CSAIL-TR-2008-055 September 13, 2008 Stohasti Combinatorial Optimization with Risk Evdokia Nikolova massahusetts institute of tehnology,

More information

Non-Orthogonal Random Access for 5G Mobile Communication Systems

Non-Orthogonal Random Access for 5G Mobile Communication Systems This artile has been aepted for publiation in a future issue of this journal, but has not been fully edited. Content may hange prior to final publiation. Citation information: DOI.9/TVT.8.86, IEEE Non-Orthogonal

More information

The Laws of Acceleration

The Laws of Acceleration The Laws of Aeleration The Relationships between Time, Veloity, and Rate of Aeleration Copyright 2001 Joseph A. Rybzyk Abstrat Presented is a theory in fundamental theoretial physis that establishes the

More information

Sensor management for PRF selection in the track-before-detect context

Sensor management for PRF selection in the track-before-detect context Sensor management for PRF seletion in the tra-before-detet ontext Fotios Katsilieris, Yvo Boers, and Hans Driessen Thales Nederland B.V. Haasbergerstraat 49, 7554 PA Hengelo, the Netherlands Email: {Fotios.Katsilieris,

More information

A simple expression for radial distribution functions of pure fluids and mixtures

A simple expression for radial distribution functions of pure fluids and mixtures A simple expression for radial distribution funtions of pure fluids and mixtures Enrio Matteoli a) Istituto di Chimia Quantistia ed Energetia Moleolare, CNR, Via Risorgimento, 35, 56126 Pisa, Italy G.

More information

LOGISTIC REGRESSION IN DEPRESSION CLASSIFICATION

LOGISTIC REGRESSION IN DEPRESSION CLASSIFICATION LOGISIC REGRESSIO I DEPRESSIO CLASSIFICAIO J. Kual,. V. ran, M. Bareš KSE, FJFI, CVU v Praze PCP, CS, 3LF UK v Praze Abstrat Well nown logisti regression and the other binary response models an be used

More information

An I-Vector Backend for Speaker Verification

An I-Vector Backend for Speaker Verification An I-Vetor Bakend for Speaker Verifiation Patrik Kenny, 1 Themos Stafylakis, 1 Jahangir Alam, 1 and Marel Kokmann 2 1 CRIM, Canada, {patrik.kenny, themos.stafylakis, jahangir.alam}@rim.a 2 VoieTrust, Canada,

More information

Multi-version Coding for Consistent Distributed Storage of Correlated Data Updates

Multi-version Coding for Consistent Distributed Storage of Correlated Data Updates Multi-version Coding for Consistent Distributed Storage of Correlated Data Updates Ramy E. Ali and Vivek R. Cadambe 1 arxiv:1708.06042v1 [s.it] 21 Aug 2017 Abstrat Motivated by appliations of distributed

More information

A Characterization of Wavelet Convergence in Sobolev Spaces

A Characterization of Wavelet Convergence in Sobolev Spaces A Charaterization of Wavelet Convergene in Sobolev Spaes Mark A. Kon 1 oston University Louise Arakelian Raphael Howard University Dediated to Prof. Robert Carroll on the oasion of his 70th birthday. Abstrat

More information

Sensitivity analysis for linear optimization problem with fuzzy data in the objective function

Sensitivity analysis for linear optimization problem with fuzzy data in the objective function Sensitivity analysis for linear optimization problem with fuzzy data in the objetive funtion Stephan Dempe, Tatiana Starostina May 5, 2004 Abstrat Linear programming problems with fuzzy oeffiients in the

More information

Seismic dip estimation based on the two-dimensional Hilbert transform and its application in random noise attenuation a

Seismic dip estimation based on the two-dimensional Hilbert transform and its application in random noise attenuation a Seismi dip estimation based on the two-dimensional Hilbert transform and its appliation in random noise attenuation a a Published in Applied Geophysis, 1, 55-63 (Marh 015) Cai Liu, Changle Chen, Dian Wang,

More information

Dynamic Rate and Channel Selection in Cognitive Radio Systems

Dynamic Rate and Channel Selection in Cognitive Radio Systems 1 Dynami Rate and Channel Seletion in Cognitive Radio Systems R. Combes and A. Proutiere KTH, The Royal Institute of Tehnology arxiv:1402.5666v2 [s.it] 12 May 2014 Abstrat In this paper, we investigate

More information

DESIGN FOR DIGITAL COMMUNICATION SYSTEMS VIA SAMPLED-DATA H CONTROL

DESIGN FOR DIGITAL COMMUNICATION SYSTEMS VIA SAMPLED-DATA H CONTROL DESIG FOR DIGITAL COMMUICATIO SYSTEMS VIA SAMPLED-DATA H COTROL M agahara 1 Y Yamamoto 2 Department of Applied Analysis and Complex Dynamial Systems Graduate Shool of Informatis Kyoto University Kyoto

More information

7 Max-Flow Problems. Business Computing and Operations Research 608

7 Max-Flow Problems. Business Computing and Operations Research 608 7 Max-Flow Problems Business Computing and Operations Researh 68 7. Max-Flow Problems In what follows, we onsider a somewhat modified problem onstellation Instead of osts of transmission, vetor now indiates

More information

Resolving RIPS Measurement Ambiguity in Maximum Likelihood Estimation

Resolving RIPS Measurement Ambiguity in Maximum Likelihood Estimation 14th International Conferene on Information Fusion Chiago, Illinois, USA, July 5-8, 011 Resolving RIPS Measurement Ambiguity in Maximum Likelihood Estimation Wenhao Li, Xuezhi Wang, and Bill Moran Shool

More information

REFINED UPPER BOUNDS FOR THE LINEAR DIOPHANTINE PROBLEM OF FROBENIUS. 1. Introduction

REFINED UPPER BOUNDS FOR THE LINEAR DIOPHANTINE PROBLEM OF FROBENIUS. 1. Introduction Version of 5/2/2003 To appear in Advanes in Applied Mathematis REFINED UPPER BOUNDS FOR THE LINEAR DIOPHANTINE PROBLEM OF FROBENIUS MATTHIAS BECK AND SHELEMYAHU ZACKS Abstrat We study the Frobenius problem:

More information

max min z i i=1 x j k s.t. j=1 x j j:i T j

max min z i i=1 x j k s.t. j=1 x j j:i T j AM 221: Advaned Optimization Spring 2016 Prof. Yaron Singer Leture 22 April 18th 1 Overview In this leture, we will study the pipage rounding tehnique whih is a deterministi rounding proedure that an be

More information

Chapter 8 Hypothesis Testing

Chapter 8 Hypothesis Testing Leture 5 for BST 63: Statistial Theory II Kui Zhang, Spring Chapter 8 Hypothesis Testing Setion 8 Introdution Definition 8 A hypothesis is a statement about a population parameter Definition 8 The two

More information

Computer Science 786S - Statistical Methods in Natural Language Processing and Data Analysis Page 1

Computer Science 786S - Statistical Methods in Natural Language Processing and Data Analysis Page 1 Computer Siene 786S - Statistial Methods in Natural Language Proessing and Data Analysis Page 1 Hypothesis Testing A statistial hypothesis is a statement about the nature of the distribution of a random

More information

Diversity Analysis of Bit-Interleaved Coded Multiple Beamforming with Orthogonal Frequency Division Multiplexing

Diversity Analysis of Bit-Interleaved Coded Multiple Beamforming with Orthogonal Frequency Division Multiplexing IEEE ICC 03 - Wireless Communiations Symposium Diversity Analysis of Bit-Interleaved Coded Multiple Beamforming with Orthogonal Frequeny Division Multiplexing Boyu Li and Ender Ayanoglu Center for Pervasive

More information

Array Design for Superresolution Direction-Finding Algorithms

Array Design for Superresolution Direction-Finding Algorithms Array Design for Superresolution Diretion-Finding Algorithms Naushad Hussein Dowlut BEng, ACGI, AMIEE Athanassios Manikas PhD, DIC, AMIEE, MIEEE Department of Eletrial Eletroni Engineering Imperial College

More information

Robust Recovery of Signals From a Structured Union of Subspaces

Robust Recovery of Signals From a Structured Union of Subspaces Robust Reovery of Signals From a Strutured Union of Subspaes 1 Yonina C. Eldar, Senior Member, IEEE and Moshe Mishali, Student Member, IEEE arxiv:87.4581v2 [nlin.cg] 3 Mar 29 Abstrat Traditional sampling

More information

On the Designs and Challenges of Practical Binary Dirty Paper Coding

On the Designs and Challenges of Practical Binary Dirty Paper Coding On the Designs and Challenges of Pratial Binary Dirty Paper Coding 04 / 08 / 2009 Gyu Bum Kyung and Chih-Chun Wang Center for Wireless Systems and Appliations Shool of Eletrial and Computer Eng. Outline

More information

MOLECULAR ORBITAL THEORY- PART I

MOLECULAR ORBITAL THEORY- PART I 5.6 Physial Chemistry Leture #24-25 MOLECULAR ORBITAL THEORY- PART I At this point, we have nearly ompleted our rash-ourse introdution to quantum mehanis and we re finally ready to deal with moleules.

More information

PoS(ISCC 2017)047. Fast Acquisition for DS/FH Spread Spectrum Signals by Using Folded Sampling Digital Receiver

PoS(ISCC 2017)047. Fast Acquisition for DS/FH Spread Spectrum Signals by Using Folded Sampling Digital Receiver Fast Aquisition for DS/FH Spread Spetrum Signals by Using Folded Sampling Digital Reeiver Beijing Institute of Satellite Information Engineering Beijing, 100086, China E-mail:ziwen7189@aliyun.om Bo Yang

More information

RESEARCH ON RANDOM FOURIER WAVE-NUMBER SPECTRUM OF FLUCTUATING WIND SPEED

RESEARCH ON RANDOM FOURIER WAVE-NUMBER SPECTRUM OF FLUCTUATING WIND SPEED The Seventh Asia-Paifi Conferene on Wind Engineering, November 8-1, 9, Taipei, Taiwan RESEARCH ON RANDOM FORIER WAVE-NMBER SPECTRM OF FLCTATING WIND SPEED Qi Yan 1, Jie Li 1 Ph D. andidate, Department

More information

On the Licensing of Innovations under Strategic Delegation

On the Licensing of Innovations under Strategic Delegation On the Liensing of Innovations under Strategi Delegation Judy Hsu Institute of Finanial Management Nanhua University Taiwan and X. Henry Wang Department of Eonomis University of Missouri USA Abstrat This

More information

Transformation to approximate independence for locally stationary Gaussian processes

Transformation to approximate independence for locally stationary Gaussian processes ransformation to approximate independene for loally stationary Gaussian proesses Joseph Guinness, Mihael L. Stein We provide new approximations for the likelihood of a time series under the loally stationary

More information

Solutions to Problem Set 1

Solutions to Problem Set 1 Eon602: Maro Theory Eonomis, HKU Instrutor: Dr. Yulei Luo September 208 Solutions to Problem Set. [0 points] Consider the following lifetime optimal onsumption-saving problem: v (a 0 ) max f;a t+ g t t

More information

Journal of Inequalities in Pure and Applied Mathematics

Journal of Inequalities in Pure and Applied Mathematics Journal of Inequalities in Pure and Applied Mathematis A NEW ARRANGEMENT INEQUALITY MOHAMMAD JAVAHERI University of Oregon Department of Mathematis Fenton Hall, Eugene, OR 97403. EMail: javaheri@uoregon.edu

More information

9 Geophysics and Radio-Astronomy: VLBI VeryLongBaseInterferometry

9 Geophysics and Radio-Astronomy: VLBI VeryLongBaseInterferometry 9 Geophysis and Radio-Astronomy: VLBI VeryLongBaseInterferometry VLBI is an interferometry tehnique used in radio astronomy, in whih two or more signals, oming from the same astronomial objet, are reeived

More information

Frequency hopping does not increase anti-jamming resilience of wireless channels

Frequency hopping does not increase anti-jamming resilience of wireless channels Frequeny hopping does not inrease anti-jamming resiliene of wireless hannels Moritz Wiese and Panos Papadimitratos Networed Systems Seurity Group KTH Royal Institute of Tehnology, Stoholm, Sweden {moritzw,

More information

Simplification of Network Dynamics in Large Systems

Simplification of Network Dynamics in Large Systems Simplifiation of Network Dynamis in Large Systems Xiaojun Lin and Ness B. Shroff Shool of Eletrial and Computer Engineering Purdue University, West Lafayette, IN 47906, U.S.A. Email: {linx, shroff}@en.purdue.edu

More information

Rigorous prediction of quadratic hyperchaotic attractors of the plane

Rigorous prediction of quadratic hyperchaotic attractors of the plane Rigorous predition of quadrati hyperhaoti attrators of the plane Zeraoulia Elhadj 1, J. C. Sprott 2 1 Department of Mathematis, University of Tébéssa, 12000), Algeria. E-mail: zeraoulia@mail.univ-tebessa.dz

More information

General Equilibrium. What happens to cause a reaction to come to equilibrium?

General Equilibrium. What happens to cause a reaction to come to equilibrium? General Equilibrium Chemial Equilibrium Most hemial reations that are enountered are reversible. In other words, they go fairly easily in either the forward or reverse diretions. The thing to remember

More information

Locally Optimal Radar Waveform Design for Detecting Doubly Spread Targets in Colored Noise

Locally Optimal Radar Waveform Design for Detecting Doubly Spread Targets in Colored Noise Preprints (www.preprints.org) NOT PEER-REVIEWED Posted: 1 Deember 17 doi:1.944/preprints171.74.v1 Artile Loally Optimal Radar Waveform Design for Deteting Doubly Spread Targets in Colored Noise Zhenghan

More information

Internet Appendix for Proxy Advisory Firms: The Economics of Selling Information to Voters

Internet Appendix for Proxy Advisory Firms: The Economics of Selling Information to Voters Internet Appendix for Proxy Advisory Firms: The Eonomis of Selling Information to Voters Andrey Malenko and Nadya Malenko The first part of the Internet Appendix presents the supplementary analysis for

More information

Lightpath routing for maximum reliability in optical mesh networks

Lightpath routing for maximum reliability in optical mesh networks Vol. 7, No. 5 / May 2008 / JOURNAL OF OPTICAL NETWORKING 449 Lightpath routing for maximum reliability in optial mesh networks Shengli Yuan, 1, * Saket Varma, 2 and Jason P. Jue 2 1 Department of Computer

More information

Average Rate Speed Scaling

Average Rate Speed Scaling Average Rate Speed Saling Nikhil Bansal David P. Bunde Ho-Leung Chan Kirk Pruhs May 2, 2008 Abstrat Speed saling is a power management tehnique that involves dynamially hanging the speed of a proessor.

More information

Broadcast Channels with Cooperating Decoders

Broadcast Channels with Cooperating Decoders To appear in the IEEE Transations on Information Theory, Deember 2006. 1 Broadast Channels with Cooperating Deoders Ron Dabora Sergio D. Servetto arxiv:s/0505032v3 [s.it] 1 Nov 2006 Abstrat We onsider

More information

Supplementary Materials

Supplementary Materials Supplementary Materials Neural population partitioning and a onurrent brain-mahine interfae for sequential motor funtion Maryam M. Shanehi, Rollin C. Hu, Marissa Powers, Gregory W. Wornell, Emery N. Brown

More information

Capacity Pooling and Cost Sharing among Independent Firms in the Presence of Congestion

Capacity Pooling and Cost Sharing among Independent Firms in the Presence of Congestion Capaity Pooling and Cost Sharing among Independent Firms in the Presene of Congestion Yimin Yu Saif Benjaafar Graduate Program in Industrial and Systems Engineering Department of Mehanial Engineering University

More information

The Lorenz Transform

The Lorenz Transform The Lorenz Transform Flameno Chuk Keyser Part I The Einstein/Bergmann deriation of the Lorentz Transform I follow the deriation of the Lorentz Transform, following Peter S Bergmann in Introdution to the

More information

SUPPORTING ultra-reliable and low-latency communications

SUPPORTING ultra-reliable and low-latency communications IEEE TRANSACTIONS ON WIRELESS COMMUNICATIONS, VOL. 17, NO. 1, JANUARY 2018 127 Cross-Layer Optimization for Ultra-Reliable and Low-Lateny Radio Aess Networs Changyang She, Member, IEEE, Chenyang Yang,

More information

Coding for Random Projections and Approximate Near Neighbor Search

Coding for Random Projections and Approximate Near Neighbor Search Coding for Random Projetions and Approximate Near Neighbor Searh Ping Li Department of Statistis & Biostatistis Department of Computer Siene Rutgers University Pisataay, NJ 8854, USA pingli@stat.rutgers.edu

More information

Error Bounds for Context Reduction and Feature Omission

Error Bounds for Context Reduction and Feature Omission Error Bounds for Context Redution and Feature Omission Eugen Bek, Ralf Shlüter, Hermann Ney,2 Human Language Tehnology and Pattern Reognition, Computer Siene Department RWTH Aahen University, Ahornstr.

More information