On Testing the Extent of Noncircularity

Size: px
Start display at page:

Download "On Testing the Extent of Noncircularity"

Transcription

1 5632 IEEE TRANSACTIONS ON SIGNAL PROCESSING, VOL. 59, NO. 11, NOVEMBER 211 On Testing the Extent of Noncircularity Mike Novey, Esa Ollila, an Tülay Aalı Abstract In this corresponence, we provie a multiple hypothesis test to etect the number of latent noncircular signals in a complex Gaussian ranom vector. Our metho sequentially tests the results of iniviual generalize likelihoo ratio test (GLRT) statistics with known asymptotic istributions to form the multiple hypothesis etector. Specifically, we are able to set a threshol yieling a precise probability of error. This test can be use to statistically etermine if a given complex observation is circular Gaussian, an if not, how many latent signals in the observation are noncircular. Simulations are use to quantify the performance of the etector as compare to a etector base on the minimum escription length (MDL) criterion. The utility of the etector is shown by applying it to a beamforming application using inepenent component analysis (ICA). Inex Terms Canonical coorinates, circularity, circularity coefficients, generalize likelihoo ratio test. I. INTRODUCTION Complex-value signals are intrinsic in many applications such as magnetic resonance imaging [1], raar [2], wireless communications [3], an in transforme ata such as the Fourier transform of image ata. Recently, the secon-orer statistics of complex ranom variables, namely the information in the commonly efine covariance an the more recently efine pseuocovariance matrices [4], [5], have been exploite in signal processing algorithms. These statistics are use to classify the signal as circular or noncircular, i.e., the signal is secon-orer circular, or proper, if the pseuocovariance matrix is zero. Depening on the circularity/noncircularity properties of the ata, ifferent classes of algorithms are use for signal processing. For noncircular ata, wiely linear minimum mean square estimation provies avantages over linear estimation [6]. Also in inepenent component analysis, specific algorithms exist to hanle the noncircular non-gaussian case such as shown in [7] an [9]. Consequently, circularity etectors have been uner active research in the recent literature [1] [13], [18] [2]. In [1] an [11], a generalize likelihoo ratio test (GLRT) statistic of circularity assuming complex normality was erive an further stuie in [11] [13]. Ajuste GLRT of circularity erive in [2] has the esirable feature of being robust to epartures from Gaussianity within complex elliptically symmetric (CES) istributions with finite fourth-orer moments. In the univariate case, circularity measures base on characteristic functions were propose [19], whereas [18] consiere GLRT of circularity uner complex generalize Gaussian istribution. Manuscript receive June 23, 21; revise January 22, 211 an April 29, 211; accepte July 16, 211. Date of publication July 29, 211; ate of current version October 12, 211. The associate eitor coorinating the review of this manuscript an approving it for publication was Dr. Arie Yereor. The work of T. Aalı an M. Novey is supporte by the NSF grants NSF-CCF an NSF-IIS M. Novey an T. Aalı are with the Department of Computer Science an Electrical Engineering, University of Marylan Baltimore County, Baltimore, MD 2125 USA ( t.aali@ieee.org; mnovey1@umbc.eu). E. Ollila is with the Department of Signal Processing an Acoustics, SMARAD CoE, Aalto University School of Science an Technology, Espoo, FI-76 Aalto, an also with the Department of Mathematical Sciences, University of Oulu, Oulu 76, Finlan ( esollila@wooster.hut.fi). Color versions of one or more of the figures in this corresponence are available online at Digital Object Ientifier 1.119/TSP Recall that a circular ranom variable (r.va.) is statistically uncorrelate with its complex conjugate. If z represents a noncircular complex ranom vector (r.v.) in, then besies the conventional covariance matrix C(z) [(z )(z ) H ], also the pseuo-covariance matrix [21] P(z) [(z )(z ) T ] nees to be calculate to obtain full secon-orer knowlege of correlations between the components. Above [z] enotes the mean of z. Ifz is circular, then P =. More generally, r.v. z with vanishing pseuo-covariance matrix is referre to as secon-orer circular [5] or proper [21]. This is equivalent to saying that circularity coefficients [7], [22], efine as the orere singular values of the coherence matrix C PC, vanish. As shown in [11], the circularity coefficients are canonical correlations between z an its conjugate z 3. Note also that circularity coefficients are invariant uner nonsingular linear transformations of the ata. Circularity etectors propose in the literature so far are for testing secon-orer circularity of z, the null hypothesis thus being H : 1 = 2 = 111 = =(,P = ) against the general (unrestricte) alternative H 1 : i 6=for at least one i(, P 6= ). Naturally, neither the circularity coefficients nor the pseuo-covariance matrix are ever exactly equal to zero for finite sample lengths an thus we seek for evience that they are statistically significantly ifferent from zero. The test statistic of the GLRT for circularity [1], [11], [13] assuming i.i.. sample z1;...; zn from a complex -variate Gaussian istribution is `n n ln (1 ^ 2 i ) (1) where ^ i ;i = 1;...; are the maximum-likelihoo estimators (MLEs) of the circularity coefficients, i.e., the orere singular values of ^C ^P(^C ) T, where ^C = 1 n n (z i z)(zi z) H an ^P = 1 n n (z i z)(zi z) T enote the MLEs of C an P an z enotes the sample mean. The test statistic has an asymptotic chi-square istribution with ( +1)egrees of freeom uner H [13], [2], i.e., (+1). 2 We note that it is recommene to use the multiplier (n ) instea of the conventional n in `n as a small sample ajustment [13, Sec. VII-B] an that more commonly `n in (1) is represente as `n = n ln[et(i ^%^% 3 )], where ^% = ^C 1 ^P. This latter form for `n is computationally simpler as it avois the explicit computation of the circularity coefficients. However, for the problem aresse in this corresponence, the former representation of the test statistic ientifie in (1) is more insightful. If the hypothesis of circularity H is rejecte, then it is of interest to stuy the extent of circularity (EOC), i.e., how many circularity coefficients are zero, or from a ifferent point of view, the extent of noncircularity (EONC), i.e., how many circularity coefficients iffer from zero. The iea is to test the null hypothesis that k circularity coefficients vanish or, that k circularity coefficients iffer from zero. The null hypothesis is given by : k > k+1 = 111 = = against the general alternative H 1 ; note that H () H. Our EONC etector sequentially forms k = ;...; 1 GLRT statistics with null hypothesis an alternative hypothesis H 1. Each GLRT statistic has a known limiting istribution, hence a confience level an threshol can be chosen. Then the first statistic that accepts the null hypothesis, correspons to the number of noncircular signals. Note that the EONC etector implements a sequential hypothesis etector X/$ IEEE

2 IEEE TRANSACTIONS ON SIGNAL PROCESSING, VOL. 59, NO. 11, NOVEMBER similar to those use in etecting the imension of a signal subspace in [14] [16], the only ifference being the efinition of eigenvalues. Another approach, for etecting the number of sources in a noisy linear moel [17], forms a sequence of binary tests eciing between the two hypotheses: most likely of the ( k 1) assumptions having more than k sources an the most likely of the k assumptions to have a maximum of k sources. However in this metho, establishing the threshol is ifficult an must be one empirically. Because of this limitation, we i not pursue this metho further. In this corresponence, first we construct a GLRT for uner the complex Gaussian assumption in Section II by extening the work of [24], [25] for maximizing the likelihoo base on the rank of the crosscorrelation matrix in canonical correlation analysis. Secon, a multiple hypothesis test is forme using the GLRT to test for the number of noncircular sources in Section III-A. To quantify the performance of our metho, we compare its performance with a metho base on the minimum escription length principle [26] in Section III-D. We show empirically that the EONC etector provies better results over the MDL metho. In Section IV, the EONC etector is applie to a beamforming application using inepenent component analysis (ICA). Using the EONC etector to partition the ata into circular an noncircular subsets, we are able to apply ifferent ICA approaches to each ata set an increase the overall performance. II. GLRT FOR Let z 2 be a -imensional complex Gaussian ranom vector an y =[z T ; z H ] T the augmente vector form an =[ T ; H ] T the augmente mean vector. The probability ensity function (pf) of z N (; C; P) is given by [4], [27], [28] where 1 p(y; C; P) = jrj exp 1 2 (y )H R 1 (y ) R = C P P 3 C (2) is the augmente covariance matrix. Assuming n inepenent an ientically istribute samples z 1 ;...; z n, the joint pf of the sample (maximize over ) becomes 1 p(y; C; P)= n jrj exp 1 2 n (y i ^) H R 1 (y i ^) = n jrj exp n 2 22n n 1 R ^R where Y 2 enotes the augmente sample matrix, ^ = [z T ; z H ] T is the augmente sample mean vector an ^R is the augmente sample covariance matrix, i.e., a matrix of the form (2) but with C an P replace by their MLEs ^C an ^P. Now recall that a GLRT ecies the general alternative H 1 against if the ratio of the likelihoo functions is greater than a threshol with the unknown parameters replace by their maximum likelihoo estimates uner each hypothesis [23]. In this case, a GLRT statistic can be written as L n(k) = p max C;P Y; ^C; ^P p Y; C; P (k) > (3) where P (k) enotes the pseuo-covariance matrix with rank r(p (k) )= k. Note that the enominator is maximizing the likelihoo function uner the restricte parameter space efine by the null hypothesis constraining the rank of the pseuo-covariance matrix equal to k. Since P is symmetric it has a special singular value ecomposition, enote Takagi factorization [8], such that P = S 1 3S T where 3 iag( 1;...; ) enotes an orere iagonal matrix with circularity coefficients i s as its iagonal elements an S is the strong-uncorrelating transform (SUT) [7], [22]. Writing P in this form emonstrates that any i = reuces the rank of P resulting in r(p) =k constraint uner. The enominator of (3) can now be formulate as max C;P p Y; C; P (k) = max P :r(p )=k = p(y; ^C; ^P) max p (Y; C; P) C i=k+1 (1 ^ 2 i ) (4) where ^ i are the MLEs for the circularity coefficients. Since the circularity coefficients are the canonical correlations [29] between z an z 3, the secon equality (4) follows using the results from [25]. For future reference we note that the numerator of (3) is simply p(y; ^C; ^P) = n j ^Rj e n = n e n j^cj n (1 ^ 2 i ) : (5) Substituting (4) into (3) an taking two times the log gives us our GLRT statistic `n(k) =2ln(L n (k)) = n ln i=k+1 (1 ^ 2 i ) > (6) where = 2 ln( ). Stanar likelihoo ratio testing theory concerning composite hypothesis testing states that the limiting istribution of the test statistic (4) uner the hypothesis is `n(k) a 2 p where 2 p enotes chi-square pf with p egrees of freeom [23]. The avantage of such a statistic is that the threshol can be set to yiel a probability of etection, an as will be shown in Section III-C, the probability of error. Since P is symmetric, there are (+1) 2 inepenent unknown complex parameters or ( +1)real parameters [2]. However, uner the null hypothesis, there are only k(2 k +1)inepenent unknown real parameters (see Appenix A). Therefore, the number of egrees of freeom is p = ( +1) k(2 k +1)=( k)( k +1)[23]. Thus, the GLRT for uner the Gaussianity assumption rejects the null hypothesis if `n(k) > 2 p;1 (7) where p;1 2 enotes the (1 )th quantile of the chi-square istribution with p egrees of freeom. The above test is asymptotically vali with level (type I error, probability of false alarm (PFA)) equal to. For example, with PFA =:5 (an =6;k =3), the rejection region is p;1 2 =21:26. Next, we investigate the valiity of the p 2 approximation to the finite sample istribution of the test statistic at small sample lengths. For this purpose, we generate N = 5 samples from the complex normal istribution N (; I; 3), = 6, with k = 3 nonzero circularity coefficients 1 =:98, 3 = :616 an 2 =:297. Let `n;[1] 111 `n;[n] enote the orere sample of `n(k) compute from N simulate samples of length n, i.e., the sample quantiles. Then, q [j] = F 1 j:5 N ; j=1;...; N, are the corresponing theoretical quantiles (where.5 in (j:5) ) is a commonly use continuity correction). Then a ( chi-square ) plot of points (q [j] ;`n;[j] ) shoul N resemble

3 5634 IEEE TRANSACTIONS ON SIGNAL PROCESSING, VOL. 59, NO. 11, NOVEMBER 211 (4) an (5) into (8) an omitting the irrelevant aitive constant terms yiels MDL(k) =n ln k (1 ^ 2 i ) + k(2 k +1)lnn: (9) Note that the MDL approach oes not have any threshol parameter but chooses the hypothesis that minimizes (9). Fig. 1. Chi-square plots when sampling from N (; I; 3) istribution with k =3nonzero circular coefficients. The vertical an horizontal lines inicate the corresponing values of =:5 an =:1 upper quantiles. a straight line through the origin with slope one. In particular, the theoretical (1 )th quantile 2 p;1 in (7) shoul be close to the corresponing sample quantile in orer for the test to retain the chosen PFA with high accuracy. For n =1an n = 5, the sample.95 (respectively,.99) quantile confirming to PFA = :5 (respectively, =:1) were an 2.84 (respectively, an 25.79), which can be contraste to theoretical quantile p;:95 2 =21:26 (respectively, p;:99 2 =26:217). Thus, when sample length increases, the sample istribution more accurately resembles the chi-square istribution which is supporte by chi-square plots epicte in Fig. 1. Goo fits to the straight line are obtaine even for n = 5, but with smaller sample length (n = 1) the fit is rather poor especially in the upper quantiles. The figure thus recommens the usage of = :5 rather than = :1 as it is better in maintaining the preetermine PFA (especially for small lengths). We wish to highlight that we can set in the simulations P =3 an C = I without loss of generality as the test statistic (an the null hypothesis) is invariant uner invertible linear transformations of the ata. A metho to generate ranom samples from N (;I; 3) is escribe in Appenix B. III. DETECTING THE NUMBER OF NONCIRCULAR SOURCES A. EONC Detector We now evise a sequential etector for the number k of noncircular sources that procees as follows: choose PFA, e.g., = :5, an iterate for k =; 1; 2;...: 1) Calculate `n(k) using (6). 2) Estimate ^k is the first value of k for which `n(k) 2 p ;1 where p k =(k)( k +1)i.e., until the null hypothesis is not rejecte. The etector above is referre to as the EONC etector. B. MDL-Base Detector For comparison purposes, we also implement a multiple hypothesis etector base on information theoretic criterion base on MDL [26], [3]. The MDL approach on moel selection is base on the moel that yiels the minimum coe length, an in the large-sample limit becomes MDL(k) =2ln max C;P p Y; C; P (k) + h ln n (8) where h = k(2k+1) is the number of free parameters of the moel; see Appenix A. The secon term in (8) is a penalty term that increases linearly with moel complexity h an sublinearly with n. Substituting C. Comments on the Selection of As shown in Section II, a choice of = :5 is able to provie an aequate false alarm rate with small sample sizes. However with large sample sizes, a lower value of may provie enhance accuracy, i.e., lower probability of error. Let us efine the probability of error as P e = P u + P o where P u is the probability of unerestimating an P o is the probability of overestimating the number of noncircular signals using the EONC etector. First let us assume the true number of noncircular signals is k, then the probability of selecting some r<kis P r u = Pr(E c \111\E c r1 \ E rj ) (1) where E i = f`n(i) 2 p ;1g enotes the event that the GLRT statistic `n(i) is below the threshol an thus woul lea to acceptance of H (i) ; the total unerestimating probability is P u = k1 i= P u.for i the unerestimating case, the alternate hypothesis is true, therefore the GLRT statistic has an asymptotic noncentral chi-square istribution `n(r) a 2 p () with noncentrality parameter [23] while the threshol is base on the central chi-square istribution. When the circularity coefficients are close to one, i.e., highly noncircular, the noncentrality parameter can be quite large as seen by examining (4). As shown in [31], the estimates ^ = [^ 1;...; ^ ] T are consistent, hence as n! 1;g(^)! p g() in probability provie that g() is a continuous function in. Thus (4), enote by g(), can be written as `n(r) nln (1 i=r+1 2 i ) for large n. Therefore, lim! 1 `n(r)! 1 resulting in P u =since E r =. The overestimating case is quite ifferent however. An overestimating error will occur if there is no unerestimation an the correct null hypothesis is rejecte, thus the probability is P = Pr(E c \111\E c kj ) Pr(E c kj )= since Pr(Er) c = 1 for r < k. The probability of error becomes P e =, consequently a low value for is esire. However, for applications where there is a small number of samples or the expecte values of the circularity coefficients are not known, then a higher value of is warrante. We show this empirically in Fig. 2 by examining the performance of the EONC etector with hypothesis H (3) = true an n = 5 while varying the circularity coefficient values. With =6, the circularity coefficients become [; ; ; ; ; ] as is ajuste from zero to one. The EONC etector s performance is quantifie by the probability of etecting the true hypothesis efine as PD Pr(H (3) jh (3) ). As escribe above, when! 1 we woul expect PD = 1 P e =1. Fig. 2 highlights the performance increases as increases as seen by the PD curve shifting to the left, i.e., better etection at lower values. However, raising increases the P e as seen by a rop in PD, specifically with =1e1. Also note that PD approaches (1 ) at values much less than one, i.e., :4 <1. In conclusion, with no a priori knowlege of the circularity coefficient values or with small sample sizes, a value of =:5 is a statistically soun choice. Also note that the MDL etector has no such ajustment the probability of error was foun to be 1e 4 in this simulation showing that the EONC provie better performance with the same P e.

4 IEEE TRANSACTIONS ON SIGNAL PROCESSING, VOL. 59, NO. 11, NOVEMBER TABLE I CONTINGENCY EONC(PFA = :5) DETECTOR FOR n = 1, 5, 15 OBTAINING COR = 464, 81, 94 AND ERR = 5293, 148, 436, RESPECTIVELY Fig. 2. Probability of etecting the correct number of noncircular signals versus circularity coefficients values for both the EONC an MDL etectors with n = 5. D. Simulations We generate 1 samples of lengths n = 1, 5, 2 from = 6 variate CN istribution N (; I; 3) with k largest circularity coefficients generate inepenently from the uniform istribution U(:1; :95) for each sample an k 2 f; 1;...;g. The PFA for the EONC etector was set to =:5. Table I reports the contingency tables of k versus the estimate number of noncircular sources ^k for the EONC etector. Note that each row sums to 1 (= number of samples for each fixe k). As expecte, for increasing n, we have larger numbers, i.e., correct estimation results, on the iagonals. Moreover, as n increases, the contingency matrix becomes closer to a iagonal matrix with smaller sprea. We use two accuracy measures to quantify the success of etection: COR = tr(q) ( +1) ERR = k= ^k= jk ^kjq k^k where Q enotes the contingency matrix an Q k^k its (k; ^k)th element. Note that COR gives the average correct estimation results whereas ERR measures the inaccuracy of the obtaine estimates; it gives more weight to those ^Q k^k the more incorrect the obtaine estimate ^k is from true k. Note that uner perfect etection, Q = I, an COR = 1 an ERR =. These figures for EONC (PFA =:5) etector for sample lengths n = 1, 5, 15 were COR = 464, 81, 94, an ERR = 5293, 148, 436, respectively, illustrating the higher accuracy as n increases. Table II reports the contingency tables of k versus the estimate number of noncircular sources ^k for the MDL etector (6). The results for sample lengths n = 1, 5, 15 were COR = 31, 67, 835, an ERR = 981, 3854, 135, respectively. The results inicate that the EONC etector provies better performance than the MDL etector as seen by comparing the COR an ERR results for both etectors. For example, the EONC provies 33% better performance than the MDL when n = 5. IV. AN APPLICATION TO ICA In this section, we emonstrate the utility of noncircularity etection with an application to blin beamforming [32] using ICA. Suppose that z follows ICA moel, z = As, where the unknown mixing matrix A is of full rank an the unobserve source vector s = [s 1 ;...;s ] T containing the statistically inepenent components is such that s 1 ;...;s k are noncircular an s k+1 ;...;s are circular. We wish to recall that the circularity coefficients i;i = 1;...; of z TABLE II CONTINGENCY MDL DETECTOR FOR n = 1, 5, 15 OBTAINING COR = 31, 67, 835, AND ERR = 981, 3854, 135, RESPECTIVELY in this case are (assuming that ICs are of finite variance) simply the (marginal) circularity coefficients of the sources s i;i =1;...;, that j [s ]j is, i = [js j ]. We can evise hybri ICA metho (HYBICA) to fin all (both the circular an noncircular sources) as follows: 1) Use the EONC etector to fin the number k of noncircular signals. 2) Pre-filter with the SUT algorithm such that y = Sz where S is the SUT. The SUT orers y by ecreasing circularity coefficients so that it can be partitione into noncircular y nc =[y 1 ;...;y k ] T an circular y c =[y k+1 ;...;y ] T vectors.

5 5636 IEEE TRANSACTIONS ON SIGNAL PROCESSING, VOL. 59, NO. 11, NOVEMBER 211 contain both the signal of interest an the interference, whereas HY- BICA performance increases ue to the partitioning of the ata. This improvement is ue to HYBICA using the optimal algorithm, C-QAM, for the BPSK sources only. Fig. 3. Performance of HYBICA, JADE2, an C-QAM algorithms versus number of BPSK (interference) sources where the total number of sources is six. 3) Estimate the noncircular sources using the ata y nc an an ICA algorithm suite to noncircular sources. If the circularity coefficients are istinct, then y nc contains the source estimates. 4) Similarly, estimate the circular sources using y c. In the blin beamforming context, A is the irection of arrival (DOA) matrix where the columns of A correspon to the spatial frequency associate with each source across the antenna array. To highlight the utility of noncircularity etection an the HYBICA algorithm, we assume that the signals of interest are binary phase shift keying (BPSK) moulate an hence noncircular. Along with the BPSK sources, we also assume the presence of several emitters of narrow-ban interference that are non-gaussian an circular in nature. Uner these assumptions, the source vector contains k BPSK an k interference sources each having a ifferent angle of arrival. Since the sources of interest are noncircular, we use the HYBICA algorithm first to partition the source vector an then process the noncircular BPSK sources separately. However since the circularity coefficients are the same for BPSK moulate signals, the complex quarature amplitue moulation algorithm (C-QAM) escribe in [33] is use instea of SUT in step 2. The C-QAM algorithm uses a cost function matching the probability mass function of BPSK signals an hence provies near-optimal performance. Simulations are run using 5 samples of = 6sources with the number of BPSK sources varie from k = ; 1;...; 5 an interfering sources k = 6; 5;...; 1. The interference signals use in the simulations are complex (bivariate) Laplacian an therefore super-gaussian. The angle of arrivals of each source is uniformly sprea between 63 egrees with the antenna element spacing 1 2 where is the wavelength. To evaluate the separation performance of the algorithms the interference to signal ratio (ISR) is use an average over the signals of interest, in this case the BPSK moulate source estimates. The ISR is calculate as ISR = 1 k p2d q=1 (v pq ) 2 max l (v pl ) 2 1 ^WA, ^W is the emixing matrix estimator, D where V =(v pq )= 1 is the set of BPSK inices foun from matching the columns of ^W with the irections of arrival of the BPSK sources. Fig. 3 epicts the ISR versus the number of BPSK an interference sources for the HYBICA, JADE2 [32], an C-QAM [33] algorithms. The figure is obtaine by averaging the results of 1 Monte Carlo runs. As seen in the figure, the HYBICA algorithm improves the performance, specifically when there are a large number of interfering signals. Both JADE2 an C-QAM performance egraes when the sources V. CONCLUSIONS AND FUTURE WORK In this corresponence, we erive a GLRT to etect the number of noncircular signals in a complex Gaussian ranom vector. We quantifie the performance of the etector empirically an showe that the etector outperforms a etector base on the MDL criterion. We then emonstrate the utility of the etector by applying it to a beamforming application using ICA. It is very likely that the propose GLRT will eteriorate severely uner violations of the Gaussian ata assumption whereas MDL may perform more reliably. However, as our ICA example illustrate, the EONC etector can provie reliable estimates even when the Gaussianity assumption oes not hol exactly. To aress this concern to some extent, we point out that also the erive GLRT for the hypothesis can be ajuste as in [2] to obtain a test that remains vali within the wie class of CES istributions with finite fourth-orer moments. To be more specific, the ajuste GLRT statistic for the hypothesis is obtaine simply by iviing the GLRT statistic `n(k) by an ajustment factor n, efine as n 1 ^ j 2+j^% j j 2 where ^% j an ^ j are the conventional sample estimators of the circularity coefficient [12]%(z j ) an the (z ) stanarize [jz j ] [jz j ] ( [jz j ]) fourth-orer moment (z j ) of the jth marginal variable z j, respectively (j =1;...;). Then, the ajuste GLRT statistic, ` `n;aj, also possess the same asymptotic 2 p-istribution uner as the nonajuste GLRT (but uner the more general assumption of sampling from an unspecifie CES istribution with finite fourthorer moments). This result follows as in [2] ue to the fact that the Corollary 1 of [34, pp. 415] applies also for hypothesis (an the respective GLRT `n(k)). For example, if we repeat the simulations of Section III-D using the exact same setting with an exception that the ata is generate from centere complex multivariate t-istribution with = 6egrees of freeom having covariance matrix I an pseuo-covariance matrix 3, enote z t 6 (; I; 3), then for n = 15, we obtaine COR an ERR values 88 an 865 (respectively, 837 an 1278) for the ajuste EONC etector (respectively, for the MDL etector). That is, the ajuste EONC etector ha more correct estimation results (i.e., larger COR value) an better accuracy (i.e., smaller ERR value) than the MDL etector. Although the propose GLRT an the EONC etector can be robustifie against non-gaussianity ata assumption, they still suffer from being nonrobust uner heavy-taile noncircular moels or in the face of outliers. One approach to obtain a robust etector is to erive a GLRT for the hypothesis uner the assumption of sampling from the multivariate extension of the generalize Gaussian ensity (GGD); thus the iea is to exten the metho in [18] to the multivariate setting. Such tests will not be pursue herein but are a subject of a separate paper. APPENDIX A NUMBER OF MODEL PARAMETERS UNDER The number of free moel parameters are etermine by the egrees of freeom of pseuo-covariance matrix P (k) with rank k. Noting that P (k) is symmetric, it can be iagonalize using Takagi factorization

6 IEEE TRANSACTIONS ON SIGNAL PROCESSING, VOL. 59, NO. 11, NOVEMBER [8], such that P (k) = VDV T, where V is unitary an D is the iagonal matrix of k nonzero singular values. Thus, the total number of parameters is 2k real parameters of V an k real parameters of D resulting in 2k + k total parameters. However, the unitary constraint of V will reuce the number of free parameters. There are k constraints ue to v H i v i =1for i =1;...;k. Also each v H i v j =for i 6= j results in two constraints, one for the real part an one for the imaginary part the result is an aitional k(k 1) constraints. Therefore, the total number of moel parameters is h =2k + k (k + k(k 1)) = k(2 k +1). APPENDIX B GENERATION OF A SAMPLE FROM CN DISTRIBUTION A complex Gaussian r.v. z with mean =, covariance C = I an pseuo-covariance matrix P = iag( 1;...; ) (i.e., z N (; I; 3)) is generate as follows: 1) Choose circularity coefficients i s from [; 1]. 2) Generate a r.v. z from circular stanar (( = ; C = I) CN istribution. 3) Then a r.v. z following N (; I; 3) istribution is obtaine as an -linear transform: z = 1 2 A + B z A B z3 where A = iag( p 1+ i ) an B = iag( p 1 i ). Naturally, to obtain n samples, one simply repeats steps 2 3 n times. To obtain a ranom eviate from N (; C; P) with P = A3A T an C = AA H where 3 is the iagonal matrix of circularity coefficients an A 1 is the strong-uncorrelating transform (SUT), then a single aitional step is neee: 4) A r.v. z following N (; C; P) istribution is obtaine as a -linear transform from z as z = + Az. Step 4 is neee, for example, in simulations with ICA applications. REFERENCES [1] V. Calhoun, T. Aali, L. Hansen, J. Larsen, an J. Pekar, ICA of functional MRI ata: An overview, presente at the 4th Int. Symp. Inepenent Component Analysis an Blin Signal Separation, Nara, Japan, Apr. 23. [2] M. Davis, P. Biigare, an D. Chang, Statistical moeling an ml parameter estimation of complex SAR imagery, in Proc. Asilomar Conf. Signals, Syst., Comput., Nov. 27, pp [3] K. Wahee an F. Salem, Blin information-theoretic multiuser etection algorithms for DS-CDMA an WCDMA ownlink systems, IEEE Trans. Neural Netw., vol. 16, no. 4, pp , Jul. 25. [4] B. Picinbono, Secon-orer complex ranom vectors an normal istributions, IEEE Trans. Signal Process., vol. 44, no. 1, pp , Oct [5] B. Picinbono, On circularity, IEEE Trans. Signal Process., vol. 42, pp , Dec [6] B. Picinbono an P. Chevalier, Wiely linear estimation with complex ata, IEEE Trans. Signal Process., vol. 43, pp , Aug [7] J. Eriksson an V. Koivunen, Complex-value ICA using secon orer statistics, presente at the m IEEE MLSP, Sao Luis, Brazil, 24. [8] R. A. Horn an C. A. Johnson, Matrix Analysis. New York: Cambrige Univ. Press, [9] M. Novey an T. Aali, On extening the complex fast ICA algorithm to noncircular sources, IEEE Trans. Signal Process., vol. 56, pp , May 28. [1] E. Ollila an V. Koivunen, Generalize complex elliptical istributions, in Proc. 3r Sensor Array Multichannel Signal Process. Workshop, Sitges, Spain, Jul. 24, pp [11] P. Schreier, L. Scharf, an A. Hanssen, A generalize likelihoo ratio test for impropriety of complex signals, IEEE Signal Process. Lett., vol. 13, pp , Jul. 26. [12] E. Ollila, On the circularity of a complex ranom variable, IEEE Signal Process. Lett., vol. 15, pp , 28. [13] A. T. Walen an P. Rubin-Delanchy, On testing for impropriety of complex-value Gaussian vectors, IEEE Trans. Signal Process., vol. 57, pp , Mar. 29. [14] A. A. Shah an D. W. Tufts, Determination of the imension of a signal subspace, in Proc. 27th Asilomar Conf., Pacific Grove, CA, 1993, vol. 2, pp [15] G. Xu an R. Roy, Detection of number of sources via exploitation of centro-symmetry property, IEEE Trans. Signal Process., vol. 42, no. 1, pp , [16] T. W. Anerson, An Introuction to Multivariate Statistical Analysis, 3r e. New York: Wiley, 23. [17] G. Bienvenu an L. Kopp, Optimality of high resolution array processing using eigensystem approach, IEEE Trans. Acoust., Speech, Signal Process., vol. 31, no. 5, pp , [18] M. Novey, T. Aali, an A. Roy, Circularity an Gaussianity etection using the complex generalize Gaussian istribution, IEEE Signal Process. Lett., vol. 16, no. 11, pp , Nov. 29. [19] J. Eriksson, E. Ollila, an V. Koivunen, Essential statistics an tools for complex ranom variables, IEEE Trans. Signal Process., vol. 58, no. 1, pp , 21. [2] E. Ollila an V. Koivunen, Ajusting the generalize likelihoo ratio test of circularity robust to non-normality, presente at the IEEE Int. Workshop Signal Process., Perugia, Italy, Jun. 29. [21] F. Neeser an J. Massey, Proper complex ranom processes with applications to information theory, IEEE Trans. Inf. Theory, vol. 39, pp , Jul [22] J. Eriksson an V. Koivunen, Complex ranom vectors an ICA moels: Ientifiability, uniqueness, an separability, IEEE Trans. Inf. Theory, vol. 52, no. 3, pp , Mar. 26. [23] S. Kay, Funamentals of Statistical Signal Processing, Detection theory. Upper Sale River, NJ: Prentice-Hall,, [24] Y. Fujikoshi, The likelihoo ratio tests for the imensionality of regression coefficients, J. Multivariate Anal., vol. 4, pp , [25] Q. T. Zhang an K. M. Wong, Information theoretic criteria for the etermination of the number of signals in spatially correlate noise, IEEE Trans. Signal Process., vol. 41, no. 4, pp , Apr [26] M. Wax an T. Kailath, Detection of signals by information theoretic criteria, in Proc. IEEE Int. Conf. Acoust., Speech, Signal Process. (ICASSP), Apr. 1985, vol. 33, pp [27] A. v.. Bos, The multivariate complex normal istribution a generalization, IEEE Trans. Inf. Theory, vol. 41, no. 2, pp , Mar [28] P. J. Schreier an L. L. Scharf, Secon-orer analysis of improper complex ranom vectors an processes, IEEE Trans. Signal Process., vol. 51, pp , Mar. 23. [29] H. Hotelling, Relations between two sets of variates, Biometrika, vol. 28, pp , [3] J. Rissanen, Moeling by the shortest ata escription, Automatica, vol. 14, pp , [31] J. P. Delmas an H. Abeia, Asymptotic istribution of circularity coefficients estimate of complex ranom variables, Signal Process., vol. 89, pp , 29. [32] J.-F. Caroso an A. Souloumiac, Blin beamforming for non-gaussian signals, Proc. Inst. Electr. Eng. Raar Signal Process., vol. 14, pp , [33] M. Novey an T. Aali, Complex fixe-point ICA algorithm for separation of QAM sources using Gaussian mixture moel, presente at the Int. Conf. Acoust., Speech, Signal Process. (ICASSP), Honolulu, HI, Apr. 27. [34] D. E. Tyler, Robustness an efficiency of scatter matrices, Biometrika, vol. 7, pp , 1983.

Publication VI. Esa Ollila On the circularity of a complex random variable. IEEE Signal Processing Letters, volume 15, pages

Publication VI. Esa Ollila On the circularity of a complex random variable. IEEE Signal Processing Letters, volume 15, pages Publication VI Esa Ollila 2008 On the circularity of a complex rom variable IEEE Signal Processing Letters, volume 15, pages 841 844 2008 Institute of Electrical Electronics Engineers (IEEE) Reprinted,

More information

A New Minimum Description Length

A New Minimum Description Length A New Minimum Description Length Soosan Beheshti, Munther A. Dahleh Laboratory for Information an Decision Systems Massachusetts Institute of Technology soosan@mit.eu,ahleh@lis.mit.eu Abstract The minimum

More information

Publication VII Institute of Electrical and Electronics Engineers (IEEE)

Publication VII Institute of Electrical and Electronics Engineers (IEEE) Publication VII Esa Ollila and Visa Koivunen. 2009. Adjusting the generalized likelihood ratio test of circularity robust to non normality. In: Proceedings of the 10th IEEE Workshop on Signal Processing

More information

Time-of-Arrival Estimation in Non-Line-Of-Sight Environments

Time-of-Arrival Estimation in Non-Line-Of-Sight Environments 2 Conference on Information Sciences an Systems, The Johns Hopkins University, March 2, 2 Time-of-Arrival Estimation in Non-Line-Of-Sight Environments Sinan Gezici, Hisashi Kobayashi an H. Vincent Poor

More information

Least-Squares Regression on Sparse Spaces

Least-Squares Regression on Sparse Spaces Least-Squares Regression on Sparse Spaces Yuri Grinberg, Mahi Milani Far, Joelle Pineau School of Computer Science McGill University Montreal, Canaa {ygrinb,mmilan1,jpineau}@cs.mcgill.ca 1 Introuction

More information

Parameter estimation: A new approach to weighting a priori information

Parameter estimation: A new approach to weighting a priori information Parameter estimation: A new approach to weighting a priori information J.L. Mea Department of Mathematics, Boise State University, Boise, ID 83725-555 E-mail: jmea@boisestate.eu Abstract. We propose a

More information

A Modification of the Jarque-Bera Test. for Normality

A Modification of the Jarque-Bera Test. for Normality Int. J. Contemp. Math. Sciences, Vol. 8, 01, no. 17, 84-85 HIKARI Lt, www.m-hikari.com http://x.oi.org/10.1988/ijcms.01.9106 A Moification of the Jarque-Bera Test for Normality Moawa El-Fallah Ab El-Salam

More information

Survey Sampling. 1 Design-based Inference. Kosuke Imai Department of Politics, Princeton University. February 19, 2013

Survey Sampling. 1 Design-based Inference. Kosuke Imai Department of Politics, Princeton University. February 19, 2013 Survey Sampling Kosuke Imai Department of Politics, Princeton University February 19, 2013 Survey sampling is one of the most commonly use ata collection methos for social scientists. We begin by escribing

More information

u!i = a T u = 0. Then S satisfies

u!i = a T u = 0. Then S satisfies Deterministic Conitions for Subspace Ientifiability from Incomplete Sampling Daniel L Pimentel-Alarcón, Nigel Boston, Robert D Nowak University of Wisconsin-Maison Abstract Consier an r-imensional subspace

More information

19 Eigenvalues, Eigenvectors, Ordinary Differential Equations, and Control

19 Eigenvalues, Eigenvectors, Ordinary Differential Equations, and Control 19 Eigenvalues, Eigenvectors, Orinary Differential Equations, an Control This section introuces eigenvalues an eigenvectors of a matrix, an iscusses the role of the eigenvalues in etermining the behavior

More information

A Review of Multiple Try MCMC algorithms for Signal Processing

A Review of Multiple Try MCMC algorithms for Signal Processing A Review of Multiple Try MCMC algorithms for Signal Processing Luca Martino Image Processing Lab., Universitat e València (Spain) Universia Carlos III e Mari, Leganes (Spain) Abstract Many applications

More information

APPLICATION of compressed sensing (CS) in radar signal

APPLICATION of compressed sensing (CS) in radar signal A Novel Joint Compressive Single Target Detection an Parameter Estimation in Raar without Signal Reconstruction Alireza Hariri, Massou Babaie-Zaeh Department of Electrical Engineering, Sharif University

More information

Robust Forward Algorithms via PAC-Bayes and Laplace Distributions. ω Q. Pr (y(ω x) < 0) = Pr A k

Robust Forward Algorithms via PAC-Bayes and Laplace Distributions. ω Q. Pr (y(ω x) < 0) = Pr A k A Proof of Lemma 2 B Proof of Lemma 3 Proof: Since the support of LL istributions is R, two such istributions are equivalent absolutely continuous with respect to each other an the ivergence is well-efine

More information

Computing Exact Confidence Coefficients of Simultaneous Confidence Intervals for Multinomial Proportions and their Functions

Computing Exact Confidence Coefficients of Simultaneous Confidence Intervals for Multinomial Proportions and their Functions Working Paper 2013:5 Department of Statistics Computing Exact Confience Coefficients of Simultaneous Confience Intervals for Multinomial Proportions an their Functions Shaobo Jin Working Paper 2013:5

More information

TEMPORAL AND TIME-FREQUENCY CORRELATION-BASED BLIND SOURCE SEPARATION METHODS. Yannick DEVILLE

TEMPORAL AND TIME-FREQUENCY CORRELATION-BASED BLIND SOURCE SEPARATION METHODS. Yannick DEVILLE TEMPORAL AND TIME-FREQUENCY CORRELATION-BASED BLIND SOURCE SEPARATION METHODS Yannick DEVILLE Université Paul Sabatier Laboratoire Acoustique, Métrologie, Instrumentation Bât. 3RB2, 8 Route e Narbonne,

More information

Influence of weight initialization on multilayer perceptron performance

Influence of weight initialization on multilayer perceptron performance Influence of weight initialization on multilayer perceptron performance M. Karouia (1,2) T. Denœux (1) R. Lengellé (1) (1) Université e Compiègne U.R.A. CNRS 817 Heuiasyc BP 649 - F-66 Compiègne ceex -

More information

Space-time Linear Dispersion Using Coordinate Interleaving

Space-time Linear Dispersion Using Coordinate Interleaving Space-time Linear Dispersion Using Coorinate Interleaving Jinsong Wu an Steven D Blostein Department of Electrical an Computer Engineering Queen s University, Kingston, Ontario, Canaa, K7L3N6 Email: wujs@ieeeorg

More information

Performance of Eigenvalue-based Signal Detectors with Known and Unknown Noise Level

Performance of Eigenvalue-based Signal Detectors with Known and Unknown Noise Level Performance of Eigenvalue-base Signal Detectors with Known an Unknown oise Level Boaz aler Weizmann Institute of Science, Israel boaz.naler@weizmann.ac.il Feerico Penna Politecnico i Torino, Italy feerico.penna@polito.it

More information

SYNCHRONOUS SEQUENTIAL CIRCUITS

SYNCHRONOUS SEQUENTIAL CIRCUITS CHAPTER SYNCHRONOUS SEUENTIAL CIRCUITS Registers an counters, two very common synchronous sequential circuits, are introuce in this chapter. Register is a igital circuit for storing information. Contents

More information

Linear First-Order Equations

Linear First-Order Equations 5 Linear First-Orer Equations Linear first-orer ifferential equations make up another important class of ifferential equations that commonly arise in applications an are relatively easy to solve (in theory)

More information

Topic 7: Convergence of Random Variables

Topic 7: Convergence of Random Variables Topic 7: Convergence of Ranom Variables Course 003, 2016 Page 0 The Inference Problem So far, our starting point has been a given probability space (S, F, P). We now look at how to generate information

More information

Optimal CDMA Signatures: A Finite-Step Approach

Optimal CDMA Signatures: A Finite-Step Approach Optimal CDMA Signatures: A Finite-Step Approach Joel A. Tropp Inst. for Comp. Engr. an Sci. (ICES) 1 University Station C000 Austin, TX 7871 jtropp@ices.utexas.eu Inerjit. S. Dhillon Dept. of Comp. Sci.

More information

Entanglement is not very useful for estimating multiple phases

Entanglement is not very useful for estimating multiple phases PHYSICAL REVIEW A 70, 032310 (2004) Entanglement is not very useful for estimating multiple phases Manuel A. Ballester* Department of Mathematics, University of Utrecht, Box 80010, 3508 TA Utrecht, The

More information

Research Article When Inflation Causes No Increase in Claim Amounts

Research Article When Inflation Causes No Increase in Claim Amounts Probability an Statistics Volume 2009, Article ID 943926, 10 pages oi:10.1155/2009/943926 Research Article When Inflation Causes No Increase in Claim Amounts Vytaras Brazauskas, 1 Bruce L. Jones, 2 an

More information

LATTICE-BASED D-OPTIMUM DESIGN FOR FOURIER REGRESSION

LATTICE-BASED D-OPTIMUM DESIGN FOR FOURIER REGRESSION The Annals of Statistics 1997, Vol. 25, No. 6, 2313 2327 LATTICE-BASED D-OPTIMUM DESIGN FOR FOURIER REGRESSION By Eva Riccomagno, 1 Rainer Schwabe 2 an Henry P. Wynn 1 University of Warwick, Technische

More information

TIME-DELAY ESTIMATION USING FARROW-BASED FRACTIONAL-DELAY FIR FILTERS: FILTER APPROXIMATION VS. ESTIMATION ERRORS

TIME-DELAY ESTIMATION USING FARROW-BASED FRACTIONAL-DELAY FIR FILTERS: FILTER APPROXIMATION VS. ESTIMATION ERRORS TIME-DEAY ESTIMATION USING FARROW-BASED FRACTIONA-DEAY FIR FITERS: FITER APPROXIMATION VS. ESTIMATION ERRORS Mattias Olsson, Håkan Johansson, an Per öwenborg Div. of Electronic Systems, Dept. of Electrical

More information

Leaving Randomness to Nature: d-dimensional Product Codes through the lens of Generalized-LDPC codes

Leaving Randomness to Nature: d-dimensional Product Codes through the lens of Generalized-LDPC codes Leaving Ranomness to Nature: -Dimensional Prouct Coes through the lens of Generalize-LDPC coes Tavor Baharav, Kannan Ramchanran Dept. of Electrical Engineering an Computer Sciences, U.C. Berkeley {tavorb,

More information

Lectures - Week 10 Introduction to Ordinary Differential Equations (ODES) First Order Linear ODEs

Lectures - Week 10 Introduction to Ordinary Differential Equations (ODES) First Order Linear ODEs Lectures - Week 10 Introuction to Orinary Differential Equations (ODES) First Orer Linear ODEs When stuying ODEs we are consiering functions of one inepenent variable, e.g., f(x), where x is the inepenent

More information

MODELLING DEPENDENCE IN INSURANCE CLAIMS PROCESSES WITH LÉVY COPULAS ABSTRACT KEYWORDS

MODELLING DEPENDENCE IN INSURANCE CLAIMS PROCESSES WITH LÉVY COPULAS ABSTRACT KEYWORDS MODELLING DEPENDENCE IN INSURANCE CLAIMS PROCESSES WITH LÉVY COPULAS BY BENJAMIN AVANZI, LUKE C. CASSAR AND BERNARD WONG ABSTRACT In this paper we investigate the potential of Lévy copulas as a tool for

More information

Improving Estimation Accuracy in Nonrandomized Response Questioning Methods by Multiple Answers

Improving Estimation Accuracy in Nonrandomized Response Questioning Methods by Multiple Answers International Journal of Statistics an Probability; Vol 6, No 5; September 207 ISSN 927-7032 E-ISSN 927-7040 Publishe by Canaian Center of Science an Eucation Improving Estimation Accuracy in Nonranomize

More information

Capacity Analysis of MIMO Systems with Unknown Channel State Information

Capacity Analysis of MIMO Systems with Unknown Channel State Information Capacity Analysis of MIMO Systems with Unknown Channel State Information Jun Zheng an Bhaskar D. Rao Dept. of Electrical an Computer Engineering University of California at San Diego e-mail: juzheng@ucs.eu,

More information

THE EFFICIENCIES OF THE SPATIAL MEDIAN AND SPATIAL SIGN COVARIANCE MATRIX FOR ELLIPTICALLY SYMMETRIC DISTRIBUTIONS

THE EFFICIENCIES OF THE SPATIAL MEDIAN AND SPATIAL SIGN COVARIANCE MATRIX FOR ELLIPTICALLY SYMMETRIC DISTRIBUTIONS THE EFFICIENCIES OF THE SPATIAL MEDIAN AND SPATIAL SIGN COVARIANCE MATRIX FOR ELLIPTICALLY SYMMETRIC DISTRIBUTIONS BY ANDREW F. MAGYAR A issertation submitte to the Grauate School New Brunswick Rutgers,

More information

under the null hypothesis, the sign test (with continuity correction) rejects H 0 when α n + n 2 2.

under the null hypothesis, the sign test (with continuity correction) rejects H 0 when α n + n 2 2. Assignment 13 Exercise 8.4 For the hypotheses consiere in Examples 8.12 an 8.13, the sign test is base on the statistic N + = #{i : Z i > 0}. Since 2 n(n + /n 1) N(0, 1) 2 uner the null hypothesis, the

More information

Linear and quadratic approximation

Linear and quadratic approximation Linear an quaratic approximation November 11, 2013 Definition: Suppose f is a function that is ifferentiable on an interval I containing the point a. The linear approximation to f at a is the linear function

More information

Multi-View Clustering via Canonical Correlation Analysis

Multi-View Clustering via Canonical Correlation Analysis Technical Report TTI-TR-2008-5 Multi-View Clustering via Canonical Correlation Analysis Kamalika Chauhuri UC San Diego Sham M. Kakae Toyota Technological Institute at Chicago ABSTRACT Clustering ata in

More information

Robustness and Perturbations of Minimal Bases

Robustness and Perturbations of Minimal Bases Robustness an Perturbations of Minimal Bases Paul Van Dooren an Froilán M Dopico December 9, 2016 Abstract Polynomial minimal bases of rational vector subspaces are a classical concept that plays an important

More information

Concentration of Measure Inequalities for Compressive Toeplitz Matrices with Applications to Detection and System Identification

Concentration of Measure Inequalities for Compressive Toeplitz Matrices with Applications to Detection and System Identification Concentration of Measure Inequalities for Compressive Toeplitz Matrices with Applications to Detection an System Ientification Borhan M Sananaji, Tyrone L Vincent, an Michael B Wakin Abstract In this paper,

More information

Optimal Signal Detection for False Track Discrimination

Optimal Signal Detection for False Track Discrimination Optimal Signal Detection for False Track Discrimination Thomas Hanselmann Darko Mušicki Dept. of Electrical an Electronic Eng. Dept. of Electrical an Electronic Eng. The University of Melbourne The University

More information

arxiv: v4 [math.pr] 27 Jul 2016

arxiv: v4 [math.pr] 27 Jul 2016 The Asymptotic Distribution of the Determinant of a Ranom Correlation Matrix arxiv:309768v4 mathpr] 7 Jul 06 AM Hanea a, & GF Nane b a Centre of xcellence for Biosecurity Risk Analysis, University of Melbourne,

More information

Thermal conductivity of graded composites: Numerical simulations and an effective medium approximation

Thermal conductivity of graded composites: Numerical simulations and an effective medium approximation JOURNAL OF MATERIALS SCIENCE 34 (999)5497 5503 Thermal conuctivity of grae composites: Numerical simulations an an effective meium approximation P. M. HUI Department of Physics, The Chinese University

More information

New Statistical Test for Quality Control in High Dimension Data Set

New Statistical Test for Quality Control in High Dimension Data Set International Journal of Applie Engineering Research ISSN 973-456 Volume, Number 6 (7) pp. 64-649 New Statistical Test for Quality Control in High Dimension Data Set Shamshuritawati Sharif, Suzilah Ismail

More information

Sharp Thresholds. Zachary Hamaker. March 15, 2010

Sharp Thresholds. Zachary Hamaker. March 15, 2010 Sharp Threshols Zachary Hamaker March 15, 2010 Abstract The Kolmogorov Zero-One law states that for tail events on infinite-imensional probability spaces, the probability must be either zero or one. Behavior

More information

This module is part of the. Memobust Handbook. on Methodology of Modern Business Statistics

This module is part of the. Memobust Handbook. on Methodology of Modern Business Statistics This moule is part of the Memobust Hanbook on Methoology of Moern Business Statistics 26 March 2014 Metho: Balance Sampling for Multi-Way Stratification Contents General section... 3 1. Summary... 3 2.

More information

Pure Further Mathematics 1. Revision Notes

Pure Further Mathematics 1. Revision Notes Pure Further Mathematics Revision Notes June 20 2 FP JUNE 20 SDB Further Pure Complex Numbers... 3 Definitions an arithmetical operations... 3 Complex conjugate... 3 Properties... 3 Complex number plane,

More information

CONFIRMATORY FACTOR ANALYSIS

CONFIRMATORY FACTOR ANALYSIS 1 CONFIRMATORY FACTOR ANALYSIS The purpose of confirmatory factor analysis (CFA) is to explain the pattern of associations among a set of observe variables in terms of a smaller number of unerlying latent

More information

Necessary and Sufficient Conditions for Sketched Subspace Clustering

Necessary and Sufficient Conditions for Sketched Subspace Clustering Necessary an Sufficient Conitions for Sketche Subspace Clustering Daniel Pimentel-Alarcón, Laura Balzano 2, Robert Nowak University of Wisconsin-Maison, 2 University of Michigan-Ann Arbor Abstract This

More information

'HVLJQ &RQVLGHUDWLRQ LQ 0DWHULDO 6HOHFWLRQ 'HVLJQ 6HQVLWLYLW\,1752'8&7,21

'HVLJQ &RQVLGHUDWLRQ LQ 0DWHULDO 6HOHFWLRQ 'HVLJQ 6HQVLWLYLW\,1752'8&7,21 Large amping in a structural material may be either esirable or unesirable, epening on the engineering application at han. For example, amping is a esirable property to the esigner concerne with limiting

More information

THE VAN KAMPEN EXPANSION FOR LINKED DUFFING LINEAR OSCILLATORS EXCITED BY COLORED NOISE

THE VAN KAMPEN EXPANSION FOR LINKED DUFFING LINEAR OSCILLATORS EXCITED BY COLORED NOISE Journal of Soun an Vibration (1996) 191(3), 397 414 THE VAN KAMPEN EXPANSION FOR LINKED DUFFING LINEAR OSCILLATORS EXCITED BY COLORED NOISE E. M. WEINSTEIN Galaxy Scientific Corporation, 2500 English Creek

More information

The Principle of Least Action

The Principle of Least Action Chapter 7. The Principle of Least Action 7.1 Force Methos vs. Energy Methos We have so far stuie two istinct ways of analyzing physics problems: force methos, basically consisting of the application of

More information

Logarithmic spurious regressions

Logarithmic spurious regressions Logarithmic spurious regressions Robert M. e Jong Michigan State University February 5, 22 Abstract Spurious regressions, i.e. regressions in which an integrate process is regresse on another integrate

More information

Closed and Open Loop Optimal Control of Buffer and Energy of a Wireless Device

Closed and Open Loop Optimal Control of Buffer and Energy of a Wireless Device Close an Open Loop Optimal Control of Buffer an Energy of a Wireless Device V. S. Borkar School of Technology an Computer Science TIFR, umbai, Inia. borkar@tifr.res.in A. A. Kherani B. J. Prabhu INRIA

More information

arxiv: v1 [physics.flu-dyn] 8 May 2014

arxiv: v1 [physics.flu-dyn] 8 May 2014 Energetics of a flui uner the Boussinesq approximation arxiv:1405.1921v1 [physics.flu-yn] 8 May 2014 Kiyoshi Maruyama Department of Earth an Ocean Sciences, National Defense Acaemy, Yokosuka, Kanagawa

More information

Lower Bounds for the Smoothed Number of Pareto optimal Solutions

Lower Bounds for the Smoothed Number of Pareto optimal Solutions Lower Bouns for the Smoothe Number of Pareto optimal Solutions Tobias Brunsch an Heiko Röglin Department of Computer Science, University of Bonn, Germany brunsch@cs.uni-bonn.e, heiko@roeglin.org Abstract.

More information

Semiclassical analysis of long-wavelength multiphoton processes: The Rydberg atom

Semiclassical analysis of long-wavelength multiphoton processes: The Rydberg atom PHYSICAL REVIEW A 69, 063409 (2004) Semiclassical analysis of long-wavelength multiphoton processes: The Ryberg atom Luz V. Vela-Arevalo* an Ronal F. Fox Center for Nonlinear Sciences an School of Physics,

More information

Situation awareness of power system based on static voltage security region

Situation awareness of power system based on static voltage security region The 6th International Conference on Renewable Power Generation (RPG) 19 20 October 2017 Situation awareness of power system base on static voltage security region Fei Xiao, Zi-Qing Jiang, Qian Ai, Ran

More information

Modeling of Dependence Structures in Risk Management and Solvency

Modeling of Dependence Structures in Risk Management and Solvency Moeling of Depenence Structures in Risk Management an Solvency University of California, Santa Barbara 0. August 007 Doreen Straßburger Structure. Risk Measurement uner Solvency II. Copulas 3. Depenent

More information

Introduction to the Vlasov-Poisson system

Introduction to the Vlasov-Poisson system Introuction to the Vlasov-Poisson system Simone Calogero 1 The Vlasov equation Consier a particle with mass m > 0. Let x(t) R 3 enote the position of the particle at time t R an v(t) = ẋ(t) = x(t)/t its

More information

Calculus of Variations

Calculus of Variations 16.323 Lecture 5 Calculus of Variations Calculus of Variations Most books cover this material well, but Kirk Chapter 4 oes a particularly nice job. x(t) x* x*+ αδx (1) x*- αδx (1) αδx (1) αδx (1) t f t

More information

SYMMETRIC KRONECKER PRODUCTS AND SEMICLASSICAL WAVE PACKETS

SYMMETRIC KRONECKER PRODUCTS AND SEMICLASSICAL WAVE PACKETS SYMMETRIC KRONECKER PRODUCTS AND SEMICLASSICAL WAVE PACKETS GEORGE A HAGEDORN AND CAROLINE LASSER Abstract We investigate the iterate Kronecker prouct of a square matrix with itself an prove an invariance

More information

Hybrid Fusion for Biometrics: Combining Score-level and Decision-level Fusion

Hybrid Fusion for Biometrics: Combining Score-level and Decision-level Fusion Hybri Fusion for Biometrics: Combining Score-level an Decision-level Fusion Qian Tao Raymon Velhuis Signals an Systems Group, University of Twente Postbus 217, 7500AE Enschee, the Netherlans {q.tao,r.n.j.velhuis}@ewi.utwente.nl

More information

Diagonalization of Matrices Dr. E. Jacobs

Diagonalization of Matrices Dr. E. Jacobs Diagonalization of Matrices Dr. E. Jacobs One of the very interesting lessons in this course is how certain algebraic techniques can be use to solve ifferential equations. The purpose of these notes is

More information

arxiv: v2 [cond-mat.stat-mech] 11 Nov 2016

arxiv: v2 [cond-mat.stat-mech] 11 Nov 2016 Noname manuscript No. (will be inserte by the eitor) Scaling properties of the number of ranom sequential asorption iterations neee to generate saturate ranom packing arxiv:607.06668v2 [con-mat.stat-mech]

More information

Modelling and simulation of dependence structures in nonlife insurance with Bernstein copulas

Modelling and simulation of dependence structures in nonlife insurance with Bernstein copulas Moelling an simulation of epenence structures in nonlife insurance with Bernstein copulas Prof. Dr. Dietmar Pfeifer Dept. of Mathematics, University of Olenburg an AON Benfiel, Hamburg Dr. Doreen Straßburger

More information

Unit #6 - Families of Functions, Taylor Polynomials, l Hopital s Rule

Unit #6 - Families of Functions, Taylor Polynomials, l Hopital s Rule Unit # - Families of Functions, Taylor Polynomials, l Hopital s Rule Some problems an solutions selecte or aapte from Hughes-Hallett Calculus. Critical Points. Consier the function f) = 54 +. b) a) Fin

More information

Lecture Introduction. 2 Examples of Measure Concentration. 3 The Johnson-Lindenstrauss Lemma. CS-621 Theory Gems November 28, 2012

Lecture Introduction. 2 Examples of Measure Concentration. 3 The Johnson-Lindenstrauss Lemma. CS-621 Theory Gems November 28, 2012 CS-6 Theory Gems November 8, 0 Lecture Lecturer: Alesaner Mąry Scribes: Alhussein Fawzi, Dorina Thanou Introuction Toay, we will briefly iscuss an important technique in probability theory measure concentration

More information

Chapter 6: Energy-Momentum Tensors

Chapter 6: Energy-Momentum Tensors 49 Chapter 6: Energy-Momentum Tensors This chapter outlines the general theory of energy an momentum conservation in terms of energy-momentum tensors, then applies these ieas to the case of Bohm's moel.

More information

UNIFYING PCA AND MULTISCALE APPROACHES TO FAULT DETECTION AND ISOLATION

UNIFYING PCA AND MULTISCALE APPROACHES TO FAULT DETECTION AND ISOLATION UNIFYING AND MULISCALE APPROACHES O FAUL DEECION AND ISOLAION Seongkyu Yoon an John F. MacGregor Dept. Chemical Engineering, McMaster University, Hamilton Ontario Canaa L8S 4L7 yoons@mcmaster.ca macgreg@mcmaster.ca

More information

Binary Discrimination Methods for High Dimensional Data with a. Geometric Representation

Binary Discrimination Methods for High Dimensional Data with a. Geometric Representation Binary Discrimination Methos for High Dimensional Data with a Geometric Representation Ay Bolivar-Cime, Luis Miguel Corova-Roriguez Universia Juárez Autónoma e Tabasco, División Acaémica e Ciencias Básicas

More information

Speaker Adaptation Based on Sparse and Low-rank Eigenphone Matrix Estimation

Speaker Adaptation Based on Sparse and Low-rank Eigenphone Matrix Estimation INTERSPEECH 2014 Speaker Aaptation Base on Sparse an Low-rank Eigenphone Matrix Estimation Wen-Lin Zhang 1, Dan Qu 1, Wei-Qiang Zhang 2, Bi-Cheng Li 1 1 Zhengzhou Information Science an Technology Institute,

More information

EVALUATING HIGHER DERIVATIVE TENSORS BY FORWARD PROPAGATION OF UNIVARIATE TAYLOR SERIES

EVALUATING HIGHER DERIVATIVE TENSORS BY FORWARD PROPAGATION OF UNIVARIATE TAYLOR SERIES MATHEMATICS OF COMPUTATION Volume 69, Number 231, Pages 1117 1130 S 0025-5718(00)01120-0 Article electronically publishe on February 17, 2000 EVALUATING HIGHER DERIVATIVE TENSORS BY FORWARD PROPAGATION

More information

A simple model for the small-strain behaviour of soils

A simple model for the small-strain behaviour of soils A simple moel for the small-strain behaviour of soils José Jorge Naer Department of Structural an Geotechnical ngineering, Polytechnic School, University of São Paulo 05508-900, São Paulo, Brazil, e-mail:

More information

Perfect Matchings in Õ(n1.5 ) Time in Regular Bipartite Graphs

Perfect Matchings in Õ(n1.5 ) Time in Regular Bipartite Graphs Perfect Matchings in Õ(n1.5 ) Time in Regular Bipartite Graphs Ashish Goel Michael Kapralov Sanjeev Khanna Abstract We consier the well-stuie problem of fining a perfect matching in -regular bipartite

More information

Lower bounds on Locality Sensitive Hashing

Lower bounds on Locality Sensitive Hashing Lower bouns on Locality Sensitive Hashing Rajeev Motwani Assaf Naor Rina Panigrahy Abstract Given a metric space (X, X ), c 1, r > 0, an p, q [0, 1], a istribution over mappings H : X N is calle a (r,

More information

Characterizing Real-Valued Multivariate Complex Polynomials and Their Symmetric Tensor Representations

Characterizing Real-Valued Multivariate Complex Polynomials and Their Symmetric Tensor Representations Characterizing Real-Value Multivariate Complex Polynomials an Their Symmetric Tensor Representations Bo JIANG Zhening LI Shuzhong ZHANG December 31, 2014 Abstract In this paper we stuy multivariate polynomial

More information

An Invariance Property of the Generalized Likelihood Ratio Test

An Invariance Property of the Generalized Likelihood Ratio Test 352 IEEE SIGNAL PROCESSING LETTERS, VOL. 10, NO. 12, DECEMBER 2003 An Invariance Property of the Generalized Likelihood Ratio Test Steven M. Kay, Fellow, IEEE, and Joseph R. Gabriel, Member, IEEE Abstract

More information

Optimal Cooperative Spectrum Sensing in Cognitive Sensor Networks

Optimal Cooperative Spectrum Sensing in Cognitive Sensor Networks Optimal Cooperative Spectrum Sensing in Cognitive Sensor Networks Hai Ngoc Pham, an Zhang, Paal E. Engelsta,,3, Tor Skeie,, Frank Eliassen, Department of Informatics, University of Oslo, Norway Simula

More information

Agmon Kolmogorov Inequalities on l 2 (Z d )

Agmon Kolmogorov Inequalities on l 2 (Z d ) Journal of Mathematics Research; Vol. 6, No. ; 04 ISSN 96-9795 E-ISSN 96-9809 Publishe by Canaian Center of Science an Eucation Agmon Kolmogorov Inequalities on l (Z ) Arman Sahovic Mathematics Department,

More information

Power Generation and Distribution via Distributed Coordination Control

Power Generation and Distribution via Distributed Coordination Control Power Generation an Distribution via Distribute Coorination Control Byeong-Yeon Kim, Kwang-Kyo Oh, an Hyo-Sung Ahn arxiv:407.4870v [math.oc] 8 Jul 204 Abstract This paper presents power coorination, power

More information

Multi-edge Optimization of Low-Density Parity-Check Codes for Joint Source-Channel Coding

Multi-edge Optimization of Low-Density Parity-Check Codes for Joint Source-Channel Coding Multi-ege Optimization of Low-Density Parity-Check Coes for Joint Source-Channel Coing H. V. Beltrão Neto an W. Henkel Jacobs University Bremen Campus Ring 1 D-28759 Bremen, Germany Email: {h.beltrao,

More information

Yang Zhang, Xinmin Wang School of Automation, Northwestern Polytechnical University, Xi an , China

Yang Zhang, Xinmin Wang School of Automation, Northwestern Polytechnical University, Xi an , China oi:1.1311/1.39.6.48 Robust Aaptive Beamforming ith Null Broaening Yang Zhang, inmin ang School of Automation, Northestern Polytechnical University, i an 717, China Abstract Aaptive beamformers are sensitive

More information

Robust Low Rank Kernel Embeddings of Multivariate Distributions

Robust Low Rank Kernel Embeddings of Multivariate Distributions Robust Low Rank Kernel Embeings of Multivariate Distributions Le Song, Bo Dai College of Computing, Georgia Institute of Technology lsong@cc.gatech.eu, boai@gatech.eu Abstract Kernel embeing of istributions

More information

State estimation for predictive maintenance using Kalman filter

State estimation for predictive maintenance using Kalman filter Reliability Engineering an System Safety 66 (1999) 29 39 www.elsevier.com/locate/ress State estimation for preictive maintenance using Kalman filter S.K. Yang, T.S. Liu* Department of Mechanical Engineering,

More information

Hyperbolic Moment Equations Using Quadrature-Based Projection Methods

Hyperbolic Moment Equations Using Quadrature-Based Projection Methods Hyperbolic Moment Equations Using Quarature-Base Projection Methos J. Koellermeier an M. Torrilhon Department of Mathematics, RWTH Aachen University, Aachen, Germany Abstract. Kinetic equations like the

More information

Bayesian Estimation of the Entropy of the Multivariate Gaussian

Bayesian Estimation of the Entropy of the Multivariate Gaussian Bayesian Estimation of the Entropy of the Multivariate Gaussian Santosh Srivastava Fre Hutchinson Cancer Research Center Seattle, WA 989, USA Email: ssrivast@fhcrc.org Maya R. Gupta Department of Electrical

More information

Gaussian processes with monotonicity information

Gaussian processes with monotonicity information Gaussian processes with monotonicity information Anonymous Author Anonymous Author Unknown Institution Unknown Institution Abstract A metho for using monotonicity information in multivariate Gaussian process

More information

arxiv: v1 [stat.co] 23 Jun 2012

arxiv: v1 [stat.co] 23 Jun 2012 Noname manuscript No. (will be inserte by the eitor) Moments Calculation For the Doubly Truncate Multivariate Normal Density Manjunath B G Stefan Wilhelm arxiv:1206.5387v1 [stat.co] 23 Jun 2012 This version:

More information

Similarity Measures for Categorical Data A Comparative Study. Technical Report

Similarity Measures for Categorical Data A Comparative Study. Technical Report Similarity Measures for Categorical Data A Comparative Stuy Technical Report Department of Computer Science an Engineering University of Minnesota 4-92 EECS Builing 200 Union Street SE Minneapolis, MN

More information

Error Floors in LDPC Codes: Fast Simulation, Bounds and Hardware Emulation

Error Floors in LDPC Codes: Fast Simulation, Bounds and Hardware Emulation Error Floors in LDPC Coes: Fast Simulation, Bouns an Harware Emulation Pamela Lee, Lara Dolecek, Zhengya Zhang, Venkat Anantharam, Borivoje Nikolic, an Martin J. Wainwright EECS Department University of

More information

LeChatelier Dynamics

LeChatelier Dynamics LeChatelier Dynamics Robert Gilmore Physics Department, Drexel University, Philaelphia, Pennsylvania 1914, USA (Date: June 12, 28, Levine Birthay Party: To be submitte.) Dynamics of the relaxation of a

More information

u t v t v t c a u t b a v t u t v t b a

u t v t v t c a u t b a v t u t v t b a Nonlinear Dynamical Systems In orer to iscuss nonlinear ynamical systems, we must first consier linear ynamical systems. Linear ynamical systems are just systems of linear equations like we have been stuying

More information

Sensors & Transducers 2015 by IFSA Publishing, S. L.

Sensors & Transducers 2015 by IFSA Publishing, S. L. Sensors & Transucers, Vol. 184, Issue 1, January 15, pp. 53-59 Sensors & Transucers 15 by IFSA Publishing, S. L. http://www.sensorsportal.com Non-invasive an Locally Resolve Measurement of Soun Velocity

More information

Applications of the Wronskian to ordinary linear differential equations

Applications of the Wronskian to ordinary linear differential equations Physics 116C Fall 2011 Applications of the Wronskian to orinary linear ifferential equations Consier a of n continuous functions y i (x) [i = 1,2,3,...,n], each of which is ifferentiable at least n times.

More information

Why Bernstein Polynomials Are Better: Fuzzy-Inspired Justification

Why Bernstein Polynomials Are Better: Fuzzy-Inspired Justification Why Bernstein Polynomials Are Better: Fuzzy-Inspire Justification Jaime Nava 1, Olga Kosheleva 2, an Vlaik Kreinovich 3 1,3 Department of Computer Science 2 Department of Teacher Eucation University of

More information

Balancing Expected and Worst-Case Utility in Contracting Models with Asymmetric Information and Pooling

Balancing Expected and Worst-Case Utility in Contracting Models with Asymmetric Information and Pooling Balancing Expecte an Worst-Case Utility in Contracting Moels with Asymmetric Information an Pooling R.B.O. erkkamp & W. van en Heuvel & A.P.M. Wagelmans Econometric Institute Report EI2018-01 9th January

More information

Technion - Computer Science Department - M.Sc. Thesis MSC Constrained Codes for Two-Dimensional Channels.

Technion - Computer Science Department - M.Sc. Thesis MSC Constrained Codes for Two-Dimensional Channels. Technion - Computer Science Department - M.Sc. Thesis MSC-2006- - 2006 Constraine Coes for Two-Dimensional Channels Keren Censor Technion - Computer Science Department - M.Sc. Thesis MSC-2006- - 2006 Technion

More information

Multi-View Clustering via Canonical Correlation Analysis

Multi-View Clustering via Canonical Correlation Analysis Keywors: multi-view learning, clustering, canonical correlation analysis Abstract Clustering ata in high-imensions is believe to be a har problem in general. A number of efficient clustering algorithms

More information

2Algebraic ONLINE PAGE PROOFS. foundations

2Algebraic ONLINE PAGE PROOFS. foundations Algebraic founations. Kick off with CAS. Algebraic skills.3 Pascal s triangle an binomial expansions.4 The binomial theorem.5 Sets of real numbers.6 Surs.7 Review . Kick off with CAS Playing lotto Using

More information

Permanent vs. Determinant

Permanent vs. Determinant Permanent vs. Determinant Frank Ban Introuction A major problem in theoretical computer science is the Permanent vs. Determinant problem. It asks: given an n by n matrix of ineterminates A = (a i,j ) an

More information

Stopping-Set Enumerator Approximations for Finite-Length Protograph LDPC Codes

Stopping-Set Enumerator Approximations for Finite-Length Protograph LDPC Codes Stopping-Set Enumerator Approximations for Finite-Length Protograph LDPC Coes Kaiann Fu an Achilleas Anastasopoulos Electrical Engineering an Computer Science Dept. University of Michigan, Ann Arbor, MI

More information

A new proof of the sharpness of the phase transition for Bernoulli percolation on Z d

A new proof of the sharpness of the phase transition for Bernoulli percolation on Z d A new proof of the sharpness of the phase transition for Bernoulli percolation on Z Hugo Duminil-Copin an Vincent Tassion October 8, 205 Abstract We provie a new proof of the sharpness of the phase transition

More information