Bayesian ML Sequence Detection for ISI Channels

Size: px
Start display at page:

Download "Bayesian ML Sequence Detection for ISI Channels"

Transcription

1 Bayesian ML Sequence Detection for ISI Cannels Jill K. Nelson Department of Electrical and Computer Engineering George Mason University Fairfax, VA Andrew C. Singer Department of Electrical and Computer Engineering University of Illinois Urbana, IL Abstract We propose a Bayesian tecnique for blind detection of coded data transmitted over a dispersive cannel. Te Bayesian maximum likeliood sequence detector views te cannel taps as stocastic quantities drawn from a known distribution and computes te probability of any transmitted sequence by averaging over te tap values. Te resulting pat metric requires memory of all previous symbols, and ence a tree-based algoritm is employed to find te most likely transmitted sequence. Simulation results sow tat te Bayesian detector can acieve bit error rates witin /4 db of te conventional known-cannel maximum likeliood (ML) sequence detector. I. INTRODUCTION A variety of communication cannels suffer from intersymbol interference, or ISI. Interference among transmitted symbols may be introduced by transmit or receive filtering, bandwidt limitations at te transmitter or receiver, or troug multi-pat propagation. Significant researc as been devoted to developing effective equalization metods for dispersive cannels. Common tecniques range from te bit-error rate (BER) optimal maximum a posteriori probability (MAP) detector to te low-complexity minimum mean-square error (MMSE) linear equalizer. In many communication systems, te parameters of te dispersive cannel may be affected by a variety of pysical and environmental conditions tat are not known a priori. As a result, receivers are often faced wit te callenge of detecting data transmitted over cannels tat are partially, or peraps entirely, unknown. Te beavior of unknown cannels may also vary wit time, tereby presenting an even more difficult problem to te receiver. One approac to equalization of an unknown cannel is to first generate an estimate of te cannel, often by transmitting training symbols, and use te cannel estimate to detect transmitted information []. Suc an approac is less desirable for time-varying cannels, since te estimate must be updated or regenerated as te cannel evolves. Alteatively, te cannel and data may be estimated concurrently to identify te most likely combination of cannel and associated data [] [5]. Since joint estimation of te cannel and data can be proibitively complex, approximate solutions to tis problem are often implemented in practice. Sesadri [4] proposes suc a tecnique based on te Viterbi algoritm. Rater tan anding te trellis on wic te Viterbi algoritm is performed, Sesadri s sceme allows an increase in te number of surviving sequences retained in te Viterbi trellis. Because te unknown cannel parameters used in te likeliood computation are estimated based on te data along te pat in question, Sesadri s algoritm falls in a class of algoritms known as per-survivor processing (PSP) tecniques [6]. Researcers ave also considered blind detection approaces, i.e., recovering te transmitted data witout generating an licit estimate of te cannel. Blind detection scemes, suc as te well-known constant modulus (CM) algoritm [7], typically employ estimates of iger order statistics (HOS) of te cannel to generate data estimates [8]. We propose an alteative blind detection sceme in wic, rater tan assuming a deterministic cannel response, we view te cannel taps as stocastic quantities drawn from a known probability distribution. Using Bayesian tecniques, we average over te unknown quantities to compute te probability of a sequence of transmitted symbols. A stacklike tree searc algoritm is used to identify te most likely sequence based on te Bayesian probability metric. Wile Bayesian approaces to blind detection ave been proposed in te past, tey ave been used witin different detection structures, suc as particle filtering and iterative Viterbi-based scemes [9], [0]. Reader and Cowley considered an approac similar to ours, employing Bayesian tecniques to generate sequence likelioods []. Rater tan using a tree-based algoritm, owever, te autors used te Viterbi algoritm, wic requires eiter a continually anding trellis or te sacrifice of information in prior data []. In te following section, we describe te system model under consideration. Te tree-based algoritm used for Bayesian ML sequence detection is described in Section 3. We ten derive te Bayesian pat metric in Section 4 and lore its empirical performance in Section 5. Conclusions are drawn in Section 6. II. SYSTEM MODEL We consider a system model, pictured in Figure, in wic te information bits are encoded using a rate R = r error-control code prior to transmission over te cannel. We assume te use of a random binary tree code in wic te bits on eac branc are independent of eac oter and of tose on oter brances. In practice, a convolutional code wit a large constraint lengt provides a reasonable approximation. Information bits are transmitted in blocks of lengt N, denoted

2 b N, yielding blocks of coded bits of lengt rn, denoted x rn, at te output of te encoder. Te encoded bits are transmitted over a lengt-l ISI cannel [n] wit additive wite Gaussian noise (AWGN) w[n] of variance σ = N 0 /. Te effect of te cannel can be modeled as L z n = k x n k + w n, () k=0 were z n denotes te cannel output at time n, and = [ 0... L ] denote te taps of te ISI cannel. For simplicity, we consider transmission of only binary pase sift keying (BPSK) encoded data. Transmission of data from larger symbol constellations is a straigtforward extension of te work presented ere. Eac block of received samples z rn serves as input to te detection and decoding block, wic uses an algoritm based on te stack decoder to estimate te data sequence b N by navigating te tree generated by te combined code and cannel. We assume tat te receiver as knowledge of te error-control code used at te transmitter, te lengt of te cannel, te variance of te AWGN, and te initial state of te encoder. b[n] Rate-R Encoder x[n] ISI Cannel w[n] z[n] Bayesian Detector and Decoder ˆb[n] Fig.. System model for Bayesian ML sequence detection. BPSK-encoded data is passed troug a rate R = error-control encoder, transmitted over r a cannel wit unknown taps, and processed by a detection and decoding block tat employs a stack-like tree searc algoritm. III. ML SEQUENCE DETECTION VIA TREE SEARCH Wen te cannel (or an estimate of te cannel) is known, ML sequence detection can be performed efficiently via te Viterbi algoritm, wic navigates a trellis and stores only te most likely pat to eac state [3]. Wen te cannel is unknown, owever, te conditional likeliood of any new branc in a pat, i.e., p ( b n b n ), is a function of all previous bits in te pat. Computation of te most likely sequence, ten, must be represented as navigation troug a tree rater tan troug a trellis and typically requires significantly iger computation per bit estimated. We use a sequential algoritm, adopted from te decoding of convolutional codes, to determine an estimate of te most likely sequence. Te stack algoritm, wic was originally developed as a metod for decoding tree codes and convolutional codes [4], navigates a tree (defined by te code) in searc of te pat wit te largest likeliood, or metric. A set of possible pats and teir associated metrics are stored in a list (or stack), and at eac iteration, te pat wit te largest metric is extended. Te stack algoritm terminates wen te most likely pat in te stack reaces a leaf of te tree, i.e., wen a complete pat containing N information bits as te largest metric of any pat in te stack. Note tat te various pats contained in te stack at a particular time are not all of te same lengt. Te algoritm may extend a pat of lengt k in one iteration and in te next iteration find te most likely pat to be one of lengt j for some j k. As a result, te pat metric must include a bias term to compensate for te differences in likeliood tat accompany a difference in pat lengt. As will be discussed, te bias term as a significant impact on te performance of te proposed detector. Te stack-like algoritm employed in te Bayesian sequential detection sceme searces troug a tree in a manner similar to tat used by te conventional stack decoder. Te tree navigated, owever, is generated by te combination of te code and te response of te cannel. Additionally, te metric used to order te pats is derived from te proposed Bayesian structure and assumes a lack of cannel knowledge. IV. DERIVATION OF THE BAYESIAN PATH METRIC To implement a stack-like algoritm we must derive an ression for te probability of a lengt-n sequence of bits, i.e., we must compute p(b n z rn, C (k) ). Here, C (k) denotes te code known at te current iteration k, i.e., te bits along all brances of te code tree tat ave been lored by te detector tus far. Since we assume tat eac block of data bits as been encoded using a random binary tree code, knowledge of te code bits on lored pats give no information about unknown code bits in unlored sections of te tree and ence simplifies derivation of te Bayesian pat metric. Note tat te cannel taps do not licitly appear in te probability ression p(b n z rn, C (k) ), as tey are viewed as stocastic quantities. We assume a known distribution for te parameters and average over te distribution to generate te probability of any sequence. A general form for te probability of a lengt-n sequence b n is derived as p(b n z rn, C (k) ) = p(b n, z rn, C (k) )d () = bn,, C(k) )p(b n,, C(k) ), d C(k) ) P(b n = ) C(k) b n ),, C (k) )p()d, were te tird equality follows from te joint independence of te information bits b n, te cannel taps, and te code known so far C (k). Eliminating C (k) ) since it is equal for all pats, we can simplify te Bayesian metric, denoted by m B (b n ), to te form m B (b n ) = P(b n ) b n,, C (k) )p()d. (3) We would like to find an integrable form for b n,, C (k) )p(), te integrand in (3). Because te cannel is unknown, and because te ISI cannel introduces memory in te received data, received samples beyond te pat of interest, i.e. z rn +, are dependent upon tose before. However, in order to develop a closed-form metric of practical complexity, we assume te two subsets of

3 te received sequence are nearly independent and approximate te overall likeliood by bn,, C(k) ) (4) p(z b n,, C (k) ) + b n,, C (k) ) = p(z bn,, C(k) ) + ). A similar approximation is used in te derivation of te pat metric for bot stack decoding of convolutional codes [5] and stack-based equalization of known ISI cannels [6]. Te form of p(z b n,, C (k) ) is simply te pat metric used for te conventional Viterbi algoritm and is given by were and p(z b n,, C (k) ) = i= p(z i b n,, C (k) ) (5) = σ (R zz [0] + T r zx T R )} (πσ ) /, R k zz[0] = r k zx = k zi, (6) i= k z i x i i L+, (7) i= k R k = ( ) ( ) x i i L+ x i T i L+. (8) i= To compute + ), we average over all possible lengtr(n n) sequences of code bits x rn +, yielding ( ) r(n n) + ) = +,x). x,} r(n n) (9) Suc a sum over r(n n) terms is impractically complex to include in te pat metric. To simplify te ression, we assume independence among te future received samples and approximate te joint likeliood by + ) rn i=+ p(z i ). (0) Suc an approximation can be justified using Massey s approac of appending a random tail to te lengt-n codeword [7]. Given te cannel tap values and te L code bits x i i L+, received sample z i as a Gaussian distribution wit mean E[z i,x i i L+ ] = T x i i L+ () and variance σ. Averaging over all lengt-l binary sequences, we obtain rn i=+ rn p(z i ) = () i=+ L πσ x,} L } σ (z i T x), wic appears in te Bayesian metric for any pat segment of lengt-n information bits. To generate an ression of reasonable complexity for practical implementation, we approximate te Gaussian mixture in () by a single Gaussian, i.e. were and p(z i ) πσ z } σz (z i µ z ), (3) µ z = E[z i ] = 0, (4) σ z = E[z i ] = σ + T. (5) We assume te cannel energy E = T, or an estimate tereof, is known at te receiver. Substituting te approximations in (4), (0), and (3) into te general Bayesian metric ression in (3) yields te result ( ) n ( ) / ( ) r(n n)/ m B (b n ) = πσ π(σ (6) + E) σ (R zz[0] + T r zx T R) } rn (σ zi p()d. + E) i=+ Given te quadratic onential form of (6), we coose te conjugate Gaussian prior over to allow for closed-form integration. To tis end, we assume te cannel taps are drawn from a pdf of te form f() = (π) L K ( ( µ ) T K ( µ ) )}, (7) were µ denotes te vector mean of te cannel taps, K denotes te covariance matrix of te cannel taps, and K denotes te determinant of K. Substituting (7) into (6) and integrating yields ) n (σ ( ) + E) r(n N) / M (8) ( m B (b n ) = were R zz [0] σ (σ ) µt K µ ( r zx σ + K µ ) T M M = K rn i=+ z i } (σ + E) ( r )} zx σ + K µ, ( ) R σ + K. (9) For te simulations and discussion presented ere, we consider scenarios in wic te cannel taps ave mean zero, equal variance σ, and are jointly independent. Under tese assumptions, te distribution over te cannel taps simplifies to } p() = (πσ T )L/ σ. (0)

4 Tese assumptions on te prior over te cannel taps can be interpreted as a minimum cannel knowledge scenario, since no prior information about (nonzero) tap values nor knowledge of cannel tap correlations is used. In cases in wic te system designer as some prior knowledge of te cannel tap values and/or correlation between te taps, tis information could be incorporated into te pdf over via a nonzero mean vector and non-diagonal correlation matrix. Using te prior over given in (0), te Bayesian pat metric for an ISI cannel simplifies to m B (b n ( ) = ) ( ) () σ L (σ + E) r(n N) R n σ σ + I / σ R zz[0] σ + ( R σ 4rT zx σ + I ) } σ r zx } rn (σ zi. + E) i=+ Te simplified pat metric takes a quadratic onential form. Note tat, wile te metric is a function of all received samples, future samples (tose associated wit information bits beyond te pat segment of interest) are normalized by te larger variance σ + E, indicating tat less information about tese samples is available since te associated transmitted bits are unknown. As simulation results will sow, te approximation yielding tis bias term underestimates te likeliood of future symbols, tereby favoring longer pats. Peraps te most interesting term in te simplified metric ( R σ + I σ ) r zx term in te onential. Tis is te r T zx term takes a form very similar to least squares estimation of te cannel taps and is te element of te metric ( in wic ) we see implicit leaing of te cannel. Te r T R zx σ + I σ r zx term is also te element of te metric tat proibits implementation of te algoritm via trellis searc. Because te algoritm does not generate licit cannel estimates but instead uses information about eac pat and te received data to generate a metric, te number of states increases by te size of te symbol alpabet (a factor of for BPSK) for eac increase in pat lengt. (If te cannel were known, te number of states would remain constant at L after L stages.) Te stack-like algoritm avoids an exaustive searc of te onentially increasing number of states by extending only te pat tat appears most likely rater tan every possible pat. V. SIMULATED PERFORMANCE OF THE BAYESIAN DETECTOR To evaluate te performance of te proposed Bayesian ML detection sceme, we simulate te detector for a tree-tap cannel wit impulse response [n] = 0.407δ[n + ] δ[n] δ[n ]. () Te performance of te Bayesian detector is simulated for block lengts of N = 0, 00, and 000 data bits. For comparison, we also simulate te performance of conventional ML sequence detection (via te Viterbi algoritm) over te cannel [n], using blocks of lengt 000. For simulation of bot te conventional ML detector and te Bayesian detection sceme, te information bits are encoded using a rate R = / (r = ) convolutional code wit generator matrix G(x) = [x + x + x + ]. (3) Figure sows te BER of te Bayesian and standard ML detectors over a range of SNR values. Te results reveal tat, for SNR between 4 and 9 db, te Bayesian ML detection sceme wit N = 000 acieves performance witin approximately / db of tat of te conventional known-cannel ML detector. It is also clear from Figure tat te BER performance of Bayesian ML sequence detection improves significantly as block size increases. Tis beavior can be lained by noting tat te Bayesian algoritm is using te received data, as well as te assumed transmitted data along eac pat, to lea te values of te cannel taps. Larger block lengts allow te detector more information from wic to lea te response of te cannel. Tis beavior is akin to tat of least-squares estimation, wic generates improved parameter estimates as te amount of data available increases. Note, owever, tat te performance for N = 00 is nearly equivalent to tat for N = 000, indicating tat te Bayesian detector can acieve low BER wit relatively small block lengt. Bit Error Rate MLSD Known Cannel Bayesian Detector, N=000 Bayesian Detector, N=00 Bayesian Detector, N= E b /N 0 (db) Fig.. Simulated performance of Bayesian ML sequence detector and conventional (Viterbi) ML sequence detector. Te cannel response is [n] = 0.407δ[n + ] δ[n] δ[n ], and te convolutional code as generator matrix G(x) = [x + x + x + ]. Additional simulations ave been conducted to lore te performance of te Bayesian sequential detector wen te cannel is essentially known. To generate suc a scenario, we let µ = and σ. Results reveal tat, even wen te cannel is known, te Bayesian detector suffers a performance loss of nearly / db wit respect to te Viterbi algoritm,

5 indicating tat te loss is likely due to a suboptimal bias term rater tan a lack of cannel knowledge. To address tis problem, we consider an alteative pat metric derivation tat avoids te assumption of independence among future received elements and te Gaussian approximation of p(z i ). Rater tan computing te pat likeliood conditioned on te entire received sequence, we consider te probability of te current pat based only on te received sequence tus far, i.e., p(b n z, C(k) ) (4) = P(bn ) p(z p(z C(k) ) bn,, C(k) )p()d (/) n σ L R σ + I = σ / p(z C(k) )(πσ ) / R zz [0] σ + ( R σ 4rT zx σ + I ) } σ r zx, were te prior distribution p() is given by (0). Te quantity p(z C (k) ) can be computed using an exteal stack wic, rater tan extending only one pat at eac stage, extends all pats in te stack. We can ten iteratively generate te conditional likelioods for all n according to p(z C(k) ) (5) = (/) n p(z x, )p()d. x C(k) Note tat te integral in (5) takes te same form as tat in (4). Akin to te work of Kun and Hagenauer [8], we can limit te size of te exteal stack to S entries (S N) to maintain a resonable complexity level. At te end of eac iteration, only te S largest stack entries are retained. Simulation results using te alteative metric given in (4) are presented in Figure 3. In addition to te conventional ML and Bayesian detectors, we also consider te performance of te PSP tecnique presented in [4], wic navigates a trellis rater tan a tree and uses LMS to update a cannel estimate for eac surviving pat. For all detectors, information blocks of lengt N = 500 were encoded using a convolutional code wit generator matrix G(x) = [ + x + x ]. Te exteal stack for te Bayesian detector was limited to 6 entries, and one pat to eac state was retained in te LMS-Viterbi sceme. At low SNR, all tree detectors sow similar bit error rates. As SNR increases, owever, te Viterbi-based sceme reaces an error floor, wile te Bayesian detector closely tracks te performance of te conventional ML detector. Te error floor results from pat eliminations made early in te data block wen te LMS cannel estimate as not converged. Because te Bayesian algoritm navigates a tree rater tan a trellis, it is able to postpone eliminating pat segments until it as more cannel information, tereby significantly improving bit error rate for te early bits in eac block. Bit Error Rate MLSD Known Cannel LMS Viterbi Algoritm Modified Bayesian Detector E b /N 0 (db) Fig. 3. Simulated performance of te conventional ML sequence detector, te Bayesian detector wit modified metric, and a Viterbi-based PSP algoritm employing LMS. Te cannel as impulse response [n] = 0.407δ[n+]+ 0.85δ[n] δ[n ], and te convolutional code as generator matrix G(x) = [ + x + x ]. VI. CONCLUSION We ave proposed a novel Bayesian tecnique for detecting data transmitted over an unknown ISI cannel. To allow implicit leaing of te cannel as te detector progresses, te most likely transmitted sequence is identified via a tree searc algoritm. Witout requiring training data, te Bayesian detector can acieve BER witin /4 db of ML sequence detection for a known cannel. Tese promising results warrant furter study of te Bayesian detector, peraps caracterizing its computational complexity (a Pareto random variable for standard stack decoders) and loring extensions to equalization of time-varying cannels. REFERENCES [] S. Haykin, Adaptive Filter Teory. Upper Saddle River, NJ: Prentice- Hall, 996. [] R. Amara and S. Marcos, A blind network of extended Kalman filters for nonstationary cannel equalization, in Proc. IEEE Int. Conf. on Acoustics, Speec, and Signal Processing, May 00, pp [3] L. M. Davis, I. B. Collings, and P. Hoeer, Joint MAP equalization and cannel estimation for frequency-selective and frequency-flat fast-fading cannels, IEEE Transactions on Communications, vol. 49, no., pp. 06 4, December 00. [4] N. Sesadri, Joint data and cannel estimation using blind trellis searc tecniques, IEEE Transactions on Communications, vol. 4, no. -4, pp , February-April 994. [5] R. Iltis, J. Synk, and K. Giridar, Bayesian algoritms for blind equalization using parallel adaptive filtering, IEEE Transactions on Communications, vol. 4, no. -4, pp , February-April 994. [6] R. Raeli, A. Polydoros, and T. Cing-Kae, Per-survivor processing: A general approac to MLSE in uncertain environments, IEEE Transactions on Communications, vol. 43, pp , February-April 995. [7] C. R. Jonson, P. Scniter, T. J. Endres, J. D. Bem, D. R. Brown, and R. A. Casas, Blind equalization using te constant modulus criterion: A review, Proceedings of te IEEE, vol. 86, no. 0, pp , October 998. [8] Z. Ding and Y. Li, Blind Equalization and Identification. New York, NY: Marcel Dekker, 00.

6 [9] J. Miguez and P. M. Djuric, Blind equalization by sequential importance sampling, in Proc. IEEE Int. Symposium on Circuits and Systems, May 00, pp. I 845 I 848. [0] X. Wang and R. Cen, Blind turbo equalization in Gaussian and impulsive noise, IEEE Transactions on Veicular Tecnology, vol. 50, no. 4, pp , July 00. [] D. Reader and W. Cowley, Blind maximum likeliood sequence detection, in Proceedings of te IEEE Inteational Conference on Acoustics, Speec, and Signal Processing, vol. 5, 996, pp [] D. Reader, Blind maximum likeliood sequence detection over fast fading communication cannels, P.D. dissertation, University of Sout Australia, August 996. [3] G. D. Foey, Maximum-likeliood sequence estimation of digital sequences in te presence of intersymbol interference, IEEE Transactions on Information Teory, vol. 8, pp , May 97. [4] R. E. Blaut, Algebraic Codes for Data Transmission. New York, NY: Cambridge University Press, 003. [5] A. J. Viterbi and J. K. Omura, Principles of Digital Communication and Coding. New York, NY: McGraw-Hill, 979. [6] F. Xiong, A. Zerik, and E. Swedyk, Sequential sequence estimation for cannels wit intersymbol interference of finite or infinite lengt, IEEE Transactions on Communications, vol. 38, no. 6, pp , June 990. [7] J. L. Massey, Variable-lengt codes and te fano metric, IEEE Transactions on Information Teory, vol. 8, no., pp , January 97. [8] C. Kun and J. Hagenauer, 8-PSK turbo equalization wit te listsequential (LISS) algoritm, in Inteational Symposium on Information Teory, June 004, p. 555.

NONLINEAR SYSTEMS IDENTIFICATION USING THE VOLTERRA MODEL. Georgeta Budura

NONLINEAR SYSTEMS IDENTIFICATION USING THE VOLTERRA MODEL. Georgeta Budura NONLINEAR SYSTEMS IDENTIFICATION USING THE VOLTERRA MODEL Georgeta Budura Politenica University of Timisoara, Faculty of Electronics and Telecommunications, Comm. Dep., georgeta.budura@etc.utt.ro Abstract:

More information

MIMO decorrelation for visible light communication based on angle optimization

MIMO decorrelation for visible light communication based on angle optimization MIMO decorrelation for visible ligt communication based on angle optimization Haiyong Zang, and Yijun Zu Citation: AIP Conference Proceedings 80, 09005 (07); View online: ttps://doi.org/0.03/.4977399 View

More information

Digital Filter Structures

Digital Filter Structures Digital Filter Structures Te convolution sum description of an LTI discrete-time system can, in principle, be used to implement te system For an IIR finite-dimensional system tis approac is not practical

More information

Characterization of Relay Channels Using the Bhattacharyya Parameter

Characterization of Relay Channels Using the Bhattacharyya Parameter Caracterization of Relay Cannels Using te Battacaryya Parameter Josepine P. K. Cu, Andrew W. Ecford, and Raviraj S. Adve Dept. of Electrical and Computer Engineering, University of Toronto, Toronto, Ontario,

More information

Multi-User Communication: Capacity, Duality, and Cooperation. Nihar Jindal Stanford University Dept. of Electrical Engineering

Multi-User Communication: Capacity, Duality, and Cooperation. Nihar Jindal Stanford University Dept. of Electrical Engineering Multi-User Communication: Capacity, Duality, and Cooperation Niar Jindal Stanford University Dept. of Electrical Engineering February 3, 004 Wireless Communication Vision Cellular Networks Wireless LAN

More information

A = h w (1) Error Analysis Physics 141

A = h w (1) Error Analysis Physics 141 Introduction In all brances of pysical science and engineering one deals constantly wit numbers wic results more or less directly from experimental observations. Experimental observations always ave inaccuracies.

More information

HOW TO DEAL WITH FFT SAMPLING INFLUENCES ON ADEV CALCULATIONS

HOW TO DEAL WITH FFT SAMPLING INFLUENCES ON ADEV CALCULATIONS HOW TO DEAL WITH FFT SAMPLING INFLUENCES ON ADEV CALCULATIONS Po-Ceng Cang National Standard Time & Frequency Lab., TL, Taiwan 1, Lane 551, Min-Tsu Road, Sec. 5, Yang-Mei, Taoyuan, Taiwan 36 Tel: 886 3

More information

Efficient algorithms for for clone items detection

Efficient algorithms for for clone items detection Efficient algoritms for for clone items detection Raoul Medina, Caroline Noyer, and Olivier Raynaud Raoul Medina, Caroline Noyer and Olivier Raynaud LIMOS - Université Blaise Pascal, Campus universitaire

More information

Partial data-dependent superimposed training based iterative channel estimation for OFDM systems over doubly selective channels

Partial data-dependent superimposed training based iterative channel estimation for OFDM systems over doubly selective channels Title Partial data-dependent superimposed training based iterative cannel estimation for OFDM systems over doubly selective cannels Autor(s) He, L; Ma, S; Wu, YC; Ng, TS Citation Te IEEE Conference and

More information

THE STURM-LIOUVILLE-TRANSFORMATION FOR THE SOLUTION OF VECTOR PARTIAL DIFFERENTIAL EQUATIONS. L. Trautmann, R. Rabenstein

THE STURM-LIOUVILLE-TRANSFORMATION FOR THE SOLUTION OF VECTOR PARTIAL DIFFERENTIAL EQUATIONS. L. Trautmann, R. Rabenstein Worksop on Transforms and Filter Banks (WTFB),Brandenburg, Germany, Marc 999 THE STURM-LIOUVILLE-TRANSFORMATION FOR THE SOLUTION OF VECTOR PARTIAL DIFFERENTIAL EQUATIONS L. Trautmann, R. Rabenstein Lerstul

More information

Joint channel estimation and data detection for OFDM systems over doubly selective channels. Title. He, L; Ma, S; Wu, YC; Ng, TS

Joint channel estimation and data detection for OFDM systems over doubly selective channels. Title. He, L; Ma, S; Wu, YC; Ng, TS Title Joint cannel estimation and data detection for OFDM systems over doubly selective cannels Autor(s) He, L; Ma, S; Wu, YC; Ng, TS Citation Te 20t IEEE International Symposium on Personal, Indoor and

More information

Continuous Stochastic Processes

Continuous Stochastic Processes Continuous Stocastic Processes Te term stocastic is often applied to penomena tat vary in time, wile te word random is reserved for penomena tat vary in space. Apart from tis distinction, te modelling

More information

A MONTE CARLO ANALYSIS OF THE EFFECTS OF COVARIANCE ON PROPAGATED UNCERTAINTIES

A MONTE CARLO ANALYSIS OF THE EFFECTS OF COVARIANCE ON PROPAGATED UNCERTAINTIES A MONTE CARLO ANALYSIS OF THE EFFECTS OF COVARIANCE ON PROPAGATED UNCERTAINTIES Ronald Ainswort Hart Scientific, American Fork UT, USA ABSTRACT Reports of calibration typically provide total combined uncertainties

More information

Polynomial Interpolation

Polynomial Interpolation Capter 4 Polynomial Interpolation In tis capter, we consider te important problem of approximatinga function fx, wose values at a set of distinct points x, x, x,, x n are known, by a polynomial P x suc

More information

Numerical Differentiation

Numerical Differentiation Numerical Differentiation Finite Difference Formulas for te first derivative (Using Taylor Expansion tecnique) (section 8.3.) Suppose tat f() = g() is a function of te variable, and tat as 0 te function

More information

Service Outage Based Power and Rate Allocation

Service Outage Based Power and Rate Allocation 1 Service Outage Based Power and Rate Allocation Jiangong Luo, Lang Lin, Roy Yates, and Predrag Spasojević Abstract Tis paper combines te concepts of ergodic capacity and capacity versus outage for fading

More information

Copyright IEEE, 13th Stat. Signal Proc. Workshop (SSP), July 2005, Bordeaux (F)

Copyright IEEE, 13th Stat. Signal Proc. Workshop (SSP), July 2005, Bordeaux (F) Copyrigt IEEE, 13t Stat. Signal Proc. Worksop SSP), 17-0 July 005, Bordeaux F) ROBUSTNESS ANALYSIS OF A GRADIENT IDENTIFICATION METHOD FOR A NONLINEAR WIENER SYSTEM Ernst Ascbacer, Markus Rupp Institute

More information

LIMITATIONS OF EULER S METHOD FOR NUMERICAL INTEGRATION

LIMITATIONS OF EULER S METHOD FOR NUMERICAL INTEGRATION LIMITATIONS OF EULER S METHOD FOR NUMERICAL INTEGRATION LAURA EVANS.. Introduction Not all differential equations can be explicitly solved for y. Tis can be problematic if we need to know te value of y

More information

New Distribution Theory for the Estimation of Structural Break Point in Mean

New Distribution Theory for the Estimation of Structural Break Point in Mean New Distribution Teory for te Estimation of Structural Break Point in Mean Liang Jiang Singapore Management University Xiaou Wang Te Cinese University of Hong Kong Jun Yu Singapore Management University

More information

Pre-Calculus Review Preemptive Strike

Pre-Calculus Review Preemptive Strike Pre-Calculus Review Preemptive Strike Attaced are some notes and one assignment wit tree parts. Tese are due on te day tat we start te pre-calculus review. I strongly suggest reading troug te notes torougly

More information

lecture 26: Richardson extrapolation

lecture 26: Richardson extrapolation 43 lecture 26: Ricardson extrapolation 35 Ricardson extrapolation, Romberg integration Trougout numerical analysis, one encounters procedures tat apply some simple approximation (eg, linear interpolation)

More information

New Streamfunction Approach for Magnetohydrodynamics

New Streamfunction Approach for Magnetohydrodynamics New Streamfunction Approac for Magnetoydrodynamics Kab Seo Kang Brooaven National Laboratory, Computational Science Center, Building 63, Room, Upton NY 973, USA. sang@bnl.gov Summary. We apply te finite

More information

THE hidden Markov model (HMM)-based parametric

THE hidden Markov model (HMM)-based parametric JOURNAL OF L A TEX CLASS FILES, VOL. 6, NO. 1, JANUARY 2007 1 Modeling Spectral Envelopes Using Restricted Boltzmann Macines and Deep Belief Networks for Statistical Parametric Speec Syntesis Zen-Hua Ling,

More information

ch (for some fixed positive number c) reaching c

ch (for some fixed positive number c) reaching c GSTF Journal of Matematics Statistics and Operations Researc (JMSOR) Vol. No. September 05 DOI 0.60/s4086-05-000-z Nonlinear Piecewise-defined Difference Equations wit Reciprocal and Cubic Terms Ramadan

More information

Energy efficiency in cognitive radio network: study of cooperative sensing using different channel sensing methods

Energy efficiency in cognitive radio network: study of cooperative sensing using different channel sensing methods Graduate Teses and Dissertations Graduate College 2015 Energy efficiency in cognitive radio network: study of cooperative sensing using different cannel sensing metods Cenxuan Cui Iowa State University

More information

Polynomial Interpolation

Polynomial Interpolation Capter 4 Polynomial Interpolation In tis capter, we consider te important problem of approximating a function f(x, wose values at a set of distinct points x, x, x 2,,x n are known, by a polynomial P (x

More information

Trust Degree Based Beamforming for Multi-Antenna Cooperative Communication Systems

Trust Degree Based Beamforming for Multi-Antenna Cooperative Communication Systems Trust Degree Based Beamforming for Multi-Antenna Cooperative Communication Systems Mojtaba Vaezi, Hazer Inaltekin, Wonjae Sin, H. Vincent Poor, and Junsan Zang* Department of Electrical Engineering, Princeton

More information

MATH745 Fall MATH745 Fall

MATH745 Fall MATH745 Fall MATH745 Fall 5 MATH745 Fall 5 INTRODUCTION WELCOME TO MATH 745 TOPICS IN NUMERICAL ANALYSIS Instructor: Dr Bartosz Protas Department of Matematics & Statistics Email: bprotas@mcmasterca Office HH 36, Ext

More information

Robust Parametric Modeling of Speech in Additive White Gaussian Noise

Robust Parametric Modeling of Speech in Additive White Gaussian Noise Journal of Signal and Information Processing, 05, 6, 99-08 Publised Online May 05 in SciRes. ttp://www.scirp.org/journal/jsip ttp://d.doi.org/0.436/jsip.05.600 Robust Parametric Modeling of Speec in Additive

More information

Fast Exact Univariate Kernel Density Estimation

Fast Exact Univariate Kernel Density Estimation Fast Exact Univariate Kernel Density Estimation David P. Hofmeyr Department of Statistics and Actuarial Science, Stellenbosc University arxiv:1806.00690v2 [stat.co] 12 Jul 2018 July 13, 2018 Abstract Tis

More information

7 Semiparametric Methods and Partially Linear Regression

7 Semiparametric Methods and Partially Linear Regression 7 Semiparametric Metods and Partially Linear Regression 7. Overview A model is called semiparametric if it is described by and were is nite-dimensional (e.g. parametric) and is in nite-dimensional (nonparametric).

More information

Te comparison of dierent models M i is based on teir relative probabilities, wic can be expressed, again using Bayes' teorem, in terms of prior probab

Te comparison of dierent models M i is based on teir relative probabilities, wic can be expressed, again using Bayes' teorem, in terms of prior probab To appear in: Advances in Neural Information Processing Systems 9, eds. M. C. Mozer, M. I. Jordan and T. Petsce. MIT Press, 997 Bayesian Model Comparison by Monte Carlo Caining David Barber D.Barber@aston.ac.uk

More information

[db]

[db] Blind Source Separation based on Second-Order Statistics wit Asymptotically Optimal Weigting Arie Yeredor Department of EE-Systems, el-aviv University P.O.Box 3900, el-aviv 69978, Israel Abstract Blind

More information

Notes on Neural Networks

Notes on Neural Networks Artificial neurons otes on eural etwors Paulo Eduardo Rauber 205 Consider te data set D {(x i y i ) i { n} x i R m y i R d } Te tas of supervised learning consists on finding a function f : R m R d tat

More information

EFFICIENT SUBSPACE DETECTION FOR HIGH-ORDER MIMO SYSTEMS. Hadi Sarieddeen Mohammad M. Mansour Ali Chehab

EFFICIENT SUBSPACE DETECTION FOR HIGH-ORDER MIMO SYSTEMS. Hadi Sarieddeen Mohammad M. Mansour Ali Chehab EFFICIENT SUBSPACE DETECTION FOR HIGH-ORDER MIMO SYSTEMS Hadi Sarieddeen Moammad M. Mansour Ali Ceab Department of Electrical and Computer Engineering American University of Beirut, Beirut 1107 00, Lebanon

More information

SECTION 3.2: DERIVATIVE FUNCTIONS and DIFFERENTIABILITY

SECTION 3.2: DERIVATIVE FUNCTIONS and DIFFERENTIABILITY (Section 3.2: Derivative Functions and Differentiability) 3.2.1 SECTION 3.2: DERIVATIVE FUNCTIONS and DIFFERENTIABILITY LEARNING OBJECTIVES Know, understand, and apply te Limit Definition of te Derivative

More information

Regularized Regression

Regularized Regression Regularized Regression David M. Blei Columbia University December 5, 205 Modern regression problems are ig dimensional, wic means tat te number of covariates p is large. In practice statisticians regularize

More information

5 Ordinary Differential Equations: Finite Difference Methods for Boundary Problems

5 Ordinary Differential Equations: Finite Difference Methods for Boundary Problems 5 Ordinary Differential Equations: Finite Difference Metods for Boundary Problems Read sections 10.1, 10.2, 10.4 Review questions 10.1 10.4, 10.8 10.9, 10.13 5.1 Introduction In te previous capters we

More information

How to Find the Derivative of a Function: Calculus 1

How to Find the Derivative of a Function: Calculus 1 Introduction How to Find te Derivative of a Function: Calculus 1 Calculus is not an easy matematics course Te fact tat you ave enrolled in suc a difficult subject indicates tat you are interested in te

More information

Convergence and Descent Properties for a Class of Multilevel Optimization Algorithms

Convergence and Descent Properties for a Class of Multilevel Optimization Algorithms Convergence and Descent Properties for a Class of Multilevel Optimization Algoritms Stepen G. Nas April 28, 2010 Abstract I present a multilevel optimization approac (termed MG/Opt) for te solution of

More information

AN ANALYSIS OF AMPLITUDE AND PERIOD OF ALTERNATING ICE LOADS ON CONICAL STRUCTURES

AN ANALYSIS OF AMPLITUDE AND PERIOD OF ALTERNATING ICE LOADS ON CONICAL STRUCTURES Ice in te Environment: Proceedings of te 1t IAHR International Symposium on Ice Dunedin, New Zealand, nd t December International Association of Hydraulic Engineering and Researc AN ANALYSIS OF AMPLITUDE

More information

arxiv: v1 [math.oc] 18 May 2018

arxiv: v1 [math.oc] 18 May 2018 Derivative-Free Optimization Algoritms based on Non-Commutative Maps * Jan Feiling,, Amelie Zeller, and Cristian Ebenbauer arxiv:805.0748v [mat.oc] 8 May 08 Institute for Systems Teory and Automatic Control,

More information

Combining functions: algebraic methods

Combining functions: algebraic methods Combining functions: algebraic metods Functions can be added, subtracted, multiplied, divided, and raised to a power, just like numbers or algebra expressions. If f(x) = x 2 and g(x) = x + 2, clearly f(x)

More information

Probabilistic Graphical Models Homework 1: Due January 29, 2014 at 4 pm

Probabilistic Graphical Models Homework 1: Due January 29, 2014 at 4 pm Probabilistic Grapical Models 10-708 Homework 1: Due January 29, 2014 at 4 pm Directions. Tis omework assignment covers te material presented in Lectures 1-3. You must complete all four problems to obtain

More information

Average Rate of Change

Average Rate of Change Te Derivative Tis can be tougt of as an attempt to draw a parallel (pysically and metaporically) between a line and a curve, applying te concept of slope to someting tat isn't actually straigt. Te slope

More information

Explicit Interleavers for a Repeat Accumulate Accumulate (RAA) code construction

Explicit Interleavers for a Repeat Accumulate Accumulate (RAA) code construction Eplicit Interleavers for a Repeat Accumulate Accumulate RAA code construction Venkatesan Gurusami Computer Science and Engineering University of Wasington Seattle, WA 98195, USA Email: venkat@csasingtonedu

More information

These errors are made from replacing an infinite process by finite one.

These errors are made from replacing an infinite process by finite one. Introduction :- Tis course examines problems tat can be solved by metods of approximation, tecniques we call numerical metods. We begin by considering some of te matematical and computational topics tat

More information

Flavius Guiaş. X(t + h) = X(t) + F (X(s)) ds.

Flavius Guiaş. X(t + h) = X(t) + F (X(s)) ds. Numerical solvers for large systems of ordinary differential equations based on te stocastic direct simulation metod improved by te and Runge Kutta principles Flavius Guiaş Abstract We present a numerical

More information

Taylor Series and the Mean Value Theorem of Derivatives

Taylor Series and the Mean Value Theorem of Derivatives 1 - Taylor Series and te Mean Value Teorem o Derivatives Te numerical solution o engineering and scientiic problems described by matematical models oten requires solving dierential equations. Dierential

More information

The total error in numerical differentiation

The total error in numerical differentiation AMS 147 Computational Metods and Applications Lecture 08 Copyrigt by Hongyun Wang, UCSC Recap: Loss of accuracy due to numerical cancellation A B 3, 3 ~10 16 In calculating te difference between A and

More information

2.11 That s So Derivative

2.11 That s So Derivative 2.11 Tat s So Derivative Introduction to Differential Calculus Just as one defines instantaneous velocity in terms of average velocity, we now define te instantaneous rate of cange of a function at a point

More information

IEOR 165 Lecture 10 Distribution Estimation

IEOR 165 Lecture 10 Distribution Estimation IEOR 165 Lecture 10 Distribution Estimation 1 Motivating Problem Consider a situation were we ave iid data x i from some unknown distribution. One problem of interest is estimating te distribution tat

More information

Termination Problems in Chemical Kinetics

Termination Problems in Chemical Kinetics Termination Problems in Cemical Kinetics Gianluigi Zavattaro and Luca Cardelli 2 Dip. Scienze dell Informazione, Università di Bologna, Italy 2 Microsoft Researc, Cambridge, UK Abstract. We consider nondeterministic

More information

Consider a function f we ll specify which assumptions we need to make about it in a minute. Let us reformulate the integral. 1 f(x) dx.

Consider a function f we ll specify which assumptions we need to make about it in a minute. Let us reformulate the integral. 1 f(x) dx. Capter 2 Integrals as sums and derivatives as differences We now switc to te simplest metods for integrating or differentiating a function from its function samples. A careful study of Taylor expansions

More information

Boosting Kernel Density Estimates: a Bias Reduction. Technique?

Boosting Kernel Density Estimates: a Bias Reduction. Technique? Boosting Kernel Density Estimates: a Bias Reduction Tecnique? Marco Di Marzio Dipartimento di Metodi Quantitativi e Teoria Economica, Università di Cieti-Pescara, Viale Pindaro 42, 65127 Pescara, Italy

More information

Spike train entropy-rate estimation using hierarchical Dirichlet process priors

Spike train entropy-rate estimation using hierarchical Dirichlet process priors publised in: Advances in Neural Information Processing Systems 26 (23), 276 284. Spike train entropy-rate estimation using ierarcical Diriclet process priors Karin Knudson Department of Matematics kknudson@mat.utexas.edu

More information

Overdispersed Variational Autoencoders

Overdispersed Variational Autoencoders Overdispersed Variational Autoencoders Harsil Sa, David Barber and Aleksandar Botev Department of Computer Science, University College London Alan Turing Institute arsil.sa.15@ucl.ac.uk, david.barber@ucl.ac.uk,

More information

Natural Language Understanding. Recap: probability, language models, and feedforward networks. Lecture 12: Recurrent Neural Networks and LSTMs

Natural Language Understanding. Recap: probability, language models, and feedforward networks. Lecture 12: Recurrent Neural Networks and LSTMs Natural Language Understanding Lecture 12: Recurrent Neural Networks and LSTMs Recap: probability, language models, and feedforward networks Simple Recurrent Networks Adam Lopez Credits: Mirella Lapata

More information

WYSE Academic Challenge 2004 Sectional Mathematics Solution Set

WYSE Academic Challenge 2004 Sectional Mathematics Solution Set WYSE Academic Callenge 00 Sectional Matematics Solution Set. Answer: B. Since te equation can be written in te form x + y, we ave a major 5 semi-axis of lengt 5 and minor semi-axis of lengt. Tis means

More information

MA455 Manifolds Solutions 1 May 2008

MA455 Manifolds Solutions 1 May 2008 MA455 Manifolds Solutions 1 May 2008 1. (i) Given real numbers a < b, find a diffeomorpism (a, b) R. Solution: For example first map (a, b) to (0, π/2) and ten map (0, π/2) diffeomorpically to R using

More information

LECTURE 14 NUMERICAL INTEGRATION. Find

LECTURE 14 NUMERICAL INTEGRATION. Find LECTURE 14 NUMERCAL NTEGRATON Find b a fxdx or b a vx ux fx ydy dx Often integration is required. However te form of fx may be suc tat analytical integration would be very difficult or impossible. Use

More information

NON STANDARD FITTED FINITE DIFFERENCE METHOD FOR SINGULAR PERTURBATION PROBLEMS USING CUBIC SPLINE

NON STANDARD FITTED FINITE DIFFERENCE METHOD FOR SINGULAR PERTURBATION PROBLEMS USING CUBIC SPLINE Global and Stocastic Analysis Vol. 4 No. 1, January 2017, 1-10 NON STANDARD FITTED FINITE DIFFERENCE METHOD FOR SINGULAR PERTURBATION PROBLEMS USING CUBIC SPLINE K. PHANEENDRA AND E. SIVA PRASAD Abstract.

More information

Bootstrap confidence intervals in nonparametric regression without an additive model

Bootstrap confidence intervals in nonparametric regression without an additive model Bootstrap confidence intervals in nonparametric regression witout an additive model Dimitris N. Politis Abstract Te problem of confidence interval construction in nonparametric regression via te bootstrap

More information

A Multiaxial Variable Amplitude Fatigue Life Prediction Method Based on a Plane Per Plane Damage Assessment

A Multiaxial Variable Amplitude Fatigue Life Prediction Method Based on a Plane Per Plane Damage Assessment American Journal of Mecanical and Industrial Engineering 28; 3(4): 47-54 ttp://www.sciencepublisinggroup.com/j/ajmie doi:.648/j.ajmie.2834.2 ISSN: 2575-679 (Print); ISSN: 2575-66 (Online) A Multiaxial

More information

Lecture 15. Interpolation II. 2 Piecewise polynomial interpolation Hermite splines

Lecture 15. Interpolation II. 2 Piecewise polynomial interpolation Hermite splines Lecture 5 Interpolation II Introduction In te previous lecture we focused primarily on polynomial interpolation of a set of n points. A difficulty we observed is tat wen n is large, our polynomial as to

More information

Fast optimal bandwidth selection for kernel density estimation

Fast optimal bandwidth selection for kernel density estimation Fast optimal bandwidt selection for kernel density estimation Vikas Candrakant Raykar and Ramani Duraiswami Dept of computer science and UMIACS, University of Maryland, CollegePark {vikas,ramani}@csumdedu

More information

arxiv: v1 [math.nt] 5 Jan 2019

arxiv: v1 [math.nt] 5 Jan 2019 POLYNOMIAL GENERALIZATIONS OF TWO-VARIABLE RAMANUJAN TYPE IDENTITIES arxiv:19010538v1 [matnt 5 Jan 019 JAMES MCLAUGHLIN AND ANDREW V SILLS Dedicated to Doron Zeilberger on te occasion of is sixtiet birtday

More information

Bounds on the Moments for an Ensemble of Random Decision Trees

Bounds on the Moments for an Ensemble of Random Decision Trees Noname manuscript No. (will be inserted by te editor) Bounds on te Moments for an Ensemble of Random Decision Trees Amit Durandar Received: Sep. 17, 2013 / Revised: Mar. 04, 2014 / Accepted: Jun. 30, 2014

More information

HARMONIC ALLOCATION TO MV CUSTOMERS IN RURAL DISTRIBUTION SYSTEMS

HARMONIC ALLOCATION TO MV CUSTOMERS IN RURAL DISTRIBUTION SYSTEMS HARMONIC ALLOCATION TO MV CUSTOMERS IN RURAL DISTRIBUTION SYSTEMS V Gosbell University of Wollongong Department of Electrical, Computer & Telecommunications Engineering, Wollongong, NSW 2522, Australia

More information

Profit Based Unit Commitment in Deregulated Electricity Markets Using A Hybrid Lagrangian Relaxation - Particle Swarm Optimization Approach

Profit Based Unit Commitment in Deregulated Electricity Markets Using A Hybrid Lagrangian Relaxation - Particle Swarm Optimization Approach Profit Based Unit Commitment in Deregulated Electricity Markets Using A Hybrid Lagrangian Relaxation - Particle Swarm Optimization Approac Adline K. Bikeri, Cristoper M. Maina and Peter K. Kiato Proceedings

More information

MAXIMUM A POSTERIORI ESTIMATION OF ROOM IMPULSE RESPONSES. Dinei Florencio and Zhengyou Zhang. Microsoft Research, Redmond, WA, USA

MAXIMUM A POSTERIORI ESTIMATION OF ROOM IMPULSE RESPONSES. Dinei Florencio and Zhengyou Zhang. Microsoft Research, Redmond, WA, USA MAXIMUM A POSTERIORI ESTIMATION OF ROOM IMPULSE RESPONSES Dinei Florencio and Zengyou Zang Microsoft Researc, Redmond, WA, USA ABSTRACT Estimating room impulse responses RIRs) as a number of applications,

More information

CORRELATION TEST OF RESIDUAL ERRORS IN FREQUENCY DOMAIN SYSTEM IDENTIFICATION

CORRELATION TEST OF RESIDUAL ERRORS IN FREQUENCY DOMAIN SYSTEM IDENTIFICATION IAC Symposium on System Identification, SYSID 006 Marc 9-3 006, Newcastle, Australia CORRELATION TEST O RESIDUAL ERRORS IN REQUENCY DOMAIN SYSTEM IDENTIICATION István Kollár *, Ri Pintelon **, Joan Scouens

More information

6. Non-uniform bending

6. Non-uniform bending . Non-uniform bending Introduction Definition A non-uniform bending is te case were te cross-section is not only bent but also seared. It is known from te statics tat in suc a case, te bending moment in

More information

An Empirical Bayesian interpretation and generalization of NL-means

An Empirical Bayesian interpretation and generalization of NL-means Computer Science Tecnical Report TR2010-934, October 2010 Courant Institute of Matematical Sciences, New York University ttp://cs.nyu.edu/web/researc/tecreports/reports.tml An Empirical Bayesian interpretation

More information

Deep Belief Network Training Improvement Using Elite Samples Minimizing Free Energy

Deep Belief Network Training Improvement Using Elite Samples Minimizing Free Energy Deep Belief Network Training Improvement Using Elite Samples Minimizing Free Energy Moammad Ali Keyvanrad a, Moammad Medi Homayounpour a a Laboratory for Intelligent Multimedia Processing (LIMP), Computer

More information

Computation of Bit-Error Rate of Coherent and Non-Coherent Detection M-Ary PSK With Gray Code in BFWA Systems

Computation of Bit-Error Rate of Coherent and Non-Coherent Detection M-Ary PSK With Gray Code in BFWA Systems Computation of Bit-Error Rate of Coherent and Non-Coherent Detection M-Ary PSK With Gray Code in BFWA Systems Department of Electrical Engineering, College of Engineering, Basrah University Basrah Iraq,

More information

LAPLACIAN MATRIX LEARNING FOR SMOOTH GRAPH SIGNAL REPRESENTATION

LAPLACIAN MATRIX LEARNING FOR SMOOTH GRAPH SIGNAL REPRESENTATION LAPLACIAN MATRIX LEARNING FOR SMOOTH GRAPH SIGNAL REPRESENTATION Xiaowen Dong, Dorina Tanou, Pascal Frossard and Pierre Vandergeynst Media Lab, MIT, USA xdong@mit.edu Signal Processing Laboratories, EPFL,

More information

EDML: A Method for Learning Parameters in Bayesian Networks

EDML: A Method for Learning Parameters in Bayesian Networks : A Metod for Learning Parameters in Bayesian Networks Artur Coi, Kaled S. Refaat and Adnan Darwice Computer Science Department University of California, Los Angeles {aycoi, krefaat, darwice}@cs.ucla.edu

More information

IN the mobile communication systems, the channel parameters

IN the mobile communication systems, the channel parameters Joint Data-Channel Estimation using the Particle Filtering on Multipath Fading Channels Tanya Bertozzi *, Didier Le Ruyet, Gilles Rigal * and Han Vu-Thien * DIGINEXT, 45 Impasse de la Draille, 3857 Aix

More information

Parameter Fitted Scheme for Singularly Perturbed Delay Differential Equations

Parameter Fitted Scheme for Singularly Perturbed Delay Differential Equations International Journal of Applied Science and Engineering 2013. 11, 4: 361-373 Parameter Fitted Sceme for Singularly Perturbed Delay Differential Equations Awoke Andargiea* and Y. N. Reddyb a b Department

More information

Optimal parameters for a hierarchical grid data structure for contact detection in arbitrarily polydisperse particle systems

Optimal parameters for a hierarchical grid data structure for contact detection in arbitrarily polydisperse particle systems Comp. Part. Mec. 04) :357 37 DOI 0.007/s4057-04-000-9 Optimal parameters for a ierarcical grid data structure for contact detection in arbitrarily polydisperse particle systems Dinant Krijgsman Vitaliy

More information

Chapter 5 FINITE DIFFERENCE METHOD (FDM)

Chapter 5 FINITE DIFFERENCE METHOD (FDM) MEE7 Computer Modeling Tecniques in Engineering Capter 5 FINITE DIFFERENCE METHOD (FDM) 5. Introduction to FDM Te finite difference tecniques are based upon approximations wic permit replacing differential

More information

Introduction to Derivatives

Introduction to Derivatives Introduction to Derivatives 5-Minute Review: Instantaneous Rates and Tangent Slope Recall te analogy tat we developed earlier First we saw tat te secant slope of te line troug te two points (a, f (a))

More information

Chapter 2 Performance Analysis of Call-Handling Processes in Buffered Cellular Wireless Networks

Chapter 2 Performance Analysis of Call-Handling Processes in Buffered Cellular Wireless Networks Capter 2 Performance Analysis of Call-Handling Processes in Buffered Cellular Wireless Networks In tis capter effective numerical computational procedures to calculate QoS (Quality of Service) metrics

More information

Expectation propagation for signal detection in flat-fading channels

Expectation propagation for signal detection in flat-fading channels Expectation propagation for signal detection in flat-fading channels Yuan Qi MIT Media Lab Cambridge, MA, 02139 USA yuanqi@media.mit.edu Thomas Minka CMU Statistics Department Pittsburgh, PA 15213 USA

More information

Mathematics 5 Worksheet 11 Geometry, Tangency, and the Derivative

Mathematics 5 Worksheet 11 Geometry, Tangency, and the Derivative Matematics 5 Workseet 11 Geometry, Tangency, and te Derivative Problem 1. Find te equation of a line wit slope m tat intersects te point (3, 9). Solution. Te equation for a line passing troug a point (x

More information

Numerical Analysis MTH603. dy dt = = (0) , y n+1. We obtain yn. Therefore. and. Copyright Virtual University of Pakistan 1

Numerical Analysis MTH603. dy dt = = (0) , y n+1. We obtain yn. Therefore. and. Copyright Virtual University of Pakistan 1 Numerical Analysis MTH60 PREDICTOR CORRECTOR METHOD Te metods presented so far are called single-step metods, were we ave seen tat te computation of y at t n+ tat is y n+ requires te knowledge of y n only.

More information

Bob Brown Math 251 Calculus 1 Chapter 3, Section 1 Completed 1 CCBC Dundalk

Bob Brown Math 251 Calculus 1 Chapter 3, Section 1 Completed 1 CCBC Dundalk Bob Brown Mat 251 Calculus 1 Capter 3, Section 1 Completed 1 Te Tangent Line Problem Te idea of a tangent line first arises in geometry in te context of a circle. But before we jump into a discussion of

More information

A FEASIBILITY STUDY OF PARTICLE FILTERS FOR MOBILE STATION RECEIVERS. Michael Lunglmayr, Martin Krueger, Mario Huemer

A FEASIBILITY STUDY OF PARTICLE FILTERS FOR MOBILE STATION RECEIVERS. Michael Lunglmayr, Martin Krueger, Mario Huemer A FEASIBILITY STUDY OF PARTICLE FILTERS FOR MOBILE STATION RECEIVERS Michael Lunglmayr, Martin Krueger, Mario Huemer Michael Lunglmayr and Martin Krueger are with Infineon Technologies AG, Munich email:

More information

MULTI-ANTENNA systems at both transmitter and receiver

MULTI-ANTENNA systems at both transmitter and receiver M. CICERONE, O. SIMEONE, N. GENG, U. SPAGNOLINI - MODAL ANALYSIS/FILTERING TO ESTIMATE TIME-VARYING MIMO-OFDM CHANNELS Modal analysis/filtering to estimate time-varying MIMO-OFDM cannels Marcello Cicerone,

More information

I. INTRODUCTION. A. Motivation

I. INTRODUCTION. A. Motivation Approximate Message Passing Algoritm wit Universal Denoising and Gaussian Mixture Learning Yanting Ma, Student Member, IEEE, Junan Zu, Student Member, IEEE, and Dror Baron, Senior Member, IEEE Abstract

More information

Learning based super-resolution land cover mapping

Learning based super-resolution land cover mapping earning based super-resolution land cover mapping Feng ing, Yiang Zang, Giles M. Foody IEEE Fellow, Xiaodong Xiuua Zang, Siming Fang, Wenbo Yun Du is work was supported in part by te National Basic Researc

More information

Volume 29, Issue 3. Existence of competitive equilibrium in economies with multi-member households

Volume 29, Issue 3. Existence of competitive equilibrium in economies with multi-member households Volume 29, Issue 3 Existence of competitive equilibrium in economies wit multi-member ouseolds Noriisa Sato Graduate Scool of Economics, Waseda University Abstract Tis paper focuses on te existence of

More information

4. The slope of the line 2x 7y = 8 is (a) 2/7 (b) 7/2 (c) 2 (d) 2/7 (e) None of these.

4. The slope of the line 2x 7y = 8 is (a) 2/7 (b) 7/2 (c) 2 (d) 2/7 (e) None of these. Mat 11. Test Form N Fall 016 Name. Instructions. Te first eleven problems are wort points eac. Te last six problems are wort 5 points eac. For te last six problems, you must use relevant metods of algebra

More information

2.8 The Derivative as a Function

2.8 The Derivative as a Function .8 Te Derivative as a Function Typically, we can find te derivative of a function f at many points of its domain: Definition. Suppose tat f is a function wic is differentiable at every point of an open

More information

SIMG Solution Set #5

SIMG Solution Set #5 SIMG-303-0033 Solution Set #5. Describe completely te state of polarization of eac of te following waves: (a) E [z,t] =ˆxE 0 cos [k 0 z ω 0 t] ŷe 0 cos [k 0 z ω 0 t] Bot components are traveling down te

More information

Complexity of Decoding Positive-Rate Primitive Reed-Solomon Codes

Complexity of Decoding Positive-Rate Primitive Reed-Solomon Codes 1 Complexity of Decoding Positive-Rate Primitive Reed-Solomon Codes Qi Ceng and Daqing Wan Abstract It as been proved tat te maximum likeliood decoding problem of Reed-Solomon codes is NP-ard. However,

More information

The Laplace equation, cylindrically or spherically symmetric case

The Laplace equation, cylindrically or spherically symmetric case Numerisce Metoden II, 7 4, und Übungen, 7 5 Course Notes, Summer Term 7 Some material and exercises Te Laplace equation, cylindrically or sperically symmetric case Electric and gravitational potential,

More information

CS522 - Partial Di erential Equations

CS522 - Partial Di erential Equations CS5 - Partial Di erential Equations Tibor Jánosi April 5, 5 Numerical Di erentiation In principle, di erentiation is a simple operation. Indeed, given a function speci ed as a closed-form formula, its

More information

estimate results from a recursive sceme tat generalizes te algoritms of Efron (967), Turnbull (976) and Li et al (997) by kernel smooting te data at e

estimate results from a recursive sceme tat generalizes te algoritms of Efron (967), Turnbull (976) and Li et al (997) by kernel smooting te data at e A kernel density estimate for interval censored data Tierry Ducesne and James E Staord y Abstract In tis paper we propose a kernel density estimate for interval-censored data It retains te simplicity andintuitive

More information