Noise constrained least mean absolute third algorithm

Size: px
Start display at page:

Download "Noise constrained least mean absolute third algorithm"

Transcription

1 Noise constrained least mean absolute third algorithm Sihai GUAN 1 Zhi LI 1 Abstract: he learning speed of an adaptie algorithm can be improed by properly constraining the cost function of the adaptie algorithm. Besides the stabilization of the NCLMF algorithm is more complicated whose stability depends solely on the input power of the adaptie filter and the NCLMF algorithm with unbounded repressors is not mean square stability een for a small alue of the step-size. So in this paper a noise ariance constrained least mean absolute third (LMA) algorithm is inestigated. he noise constrained LMA (NCLMA) algorithm is obtained by constraining the cost function of the standard LMA algorithm to the third-order moment of the additie noise. And it can eliate a ariety of non-gaussian distribution of noise such as Rayleigh noise Binary noise and so on. he NCLMA algorithm is a type of ariable step-size LMA algorithm where the step-size rule arises naturally from the constraints. he main aim of this wor is first time to derie the NCLMA adaptie algorithm analyze its conergence behaior mean square error (MSE) mean-square deiation () and assess its performance in different noise enironments. Finally the eperimental results in system identification applications presented here illustrate the principle and efficiency of the NCLMA algorithm. Key words: Adaptie filter NCLMF noise constrained non-gaussian NCLMA. 1. Introduction Adaptie filter (AF) algorithm is frequently employed in equalization actie noise control acoustic echo cancellation biomedical engineering and many other fields [1][]. In many communication applications additie white Gaussian noise is an appropriate model for thermal and possibly interference noise and the noise power can be accurately estimated prior to channel tap estimation. Hence the inestigation of the benefits of incorporating noise ariance nowledge into an adaptie channel estimation algorithm is of some practical interest in recent decades noise as a constraint has been studied for eample the noise constrained LMS (NCLMS) algorithms [3][]. Besides although the NCLMS-type algorithms hae many adantages we may hae worse choice for system identification with noise constraint in some case where the measurement noise is non-gaussian [5]. So the noise constrained LMF (NCLMF) algorithm was proposed [6]. But the stabilization of the LMF algorithm is more complicated than that of the LMS algorithm whose stability for a gien step-size depends solely on the input power of the adaptie filter and the LMF algorithm with unbounded repressors is not mean square stability een for a small alue of the stepsize [7]. Significantly the LMA algorithm is built on the imization of the mean of the absolute error alue to the third power. he error function is a perfect cone function with respect to filter coefficients so there is no local imum for the LMA algorithm [8]. Zhi Li @163.com Sihai Guan gcihey@sina.cn 1 School of Electro-Mechanical Engineering Xidian Uniersity N0. South aibai Road Xi an China

2 Using a constraint would mae the step size in the LMA algorithm act as a ariable step size. For the research of ariable step-size LMA algorithm Guan S and Li Z proposed an NVSLMA algorithm [9]. In the NVSLMA algorithm fewer parameters need to be set. But this algorithm not used the additie noise ariance. Ultimately on the basis of wor in [6][8] we propose a noise constrained LMA (NCLMA) algorithm by constraining the cost function of the standard LMA algorithm to the third-order moment of the additie noise ariance. Besides the mean conergence steady state MSE and of the NCLMA algorithm are deried. he computational compleity of the NCLMA algorithm is analyzed theoretically. Besides its characteristics are ealuated in different noise enironments. In order to further illustrate that the NCLMA algorithm is superior to the LMA and NCLMF algorithms a number of simulation results are carried out to corroborate the theoretical findings and as epected improed performance is obtained through the use of this technique. he paper is organized as follows: the proposed NCLMA algorithm is introduced in Section. he performance of the NCLMA algorithm is eplored in Section 3. he numerical simulations are carried out in Section and the conclusion is drawn in Section 5.. he proposed NCLMA algorithm he output of an FIR channel in the presence of an additie noise can be written as d w (1) opt where w opt w0 w1 w corresponds to a channel/system impulse response with N N1 taps. N is the filter length. he input data ector of the unnown system at time instant and the obsered output signal are denoted by ( ) ( 1) ( 1) N and respectiely. is a zero-mean i.i.d. process with an arbitrary probability density function. It has a ariance of. We consider real-alued quantities for simplicity. So e d w () he cost function to be imized is gien by m E 0 w e for m (3) Minimization of the cost function (3) oer gies the optimal weight alue at time i.e. w wopt with w opt where is the alue of the cost function (3) ealuated at w w opt. Consequently the Lagrangian for this problem can be set up as w w 1 w w () he critical alues of Eq. () are w wopt for any γ. his situation may cause conergence problem. So a new augmented Lagrangian as follows: w w w (5) he uses of the γ and γλ in Eq. (5) ensure the unique critical alue of w is w w 0. In Eq. (5) the augmented Lagrangian is imized with respect to the opt d

3 weight w and maimized with respect to λ respectiely. By applying the Robbnis Munro method [3] the weight and λ are updated as follows: 1 w where α and β are the positie learning parameters. he cost function to be imized is gien by w w w (6) 1 w (7) 3 w E e (8) Substituting the results for deriaties in Eq. (6) and Eq. (7) and redefining α and β (replace α by α/3 and βγ by β/) we obtain the following update recursions: where sgn e 1 sgn e w w e (9) 1 (10) e (11) denotes the sign function of the ariable i.e. if then otherwise sgn e 1.. α β and γ are positie parameters proiding more control oer the performance of the proposed algorithm. As suggested in [3] after fiing the noal step-size α γ should be chosen as a alue as large as possible to obtain a fast conergence rate. In additions β should be chosen as a small alue to achiee a desired MSE and. sgn 1 e 3 If 0 in (11) as e we get what we call the zero noise 1 1 constrained LMA (ZNCLMA) algorithm which is similar to the NCLMA but results in a mismatch due to the fact that the true noise absolute third moment is replaced by zero. he ZNCLMA can be analyzed using the same approach used to analyze the NCLMA in the sequel. Also it can be obsered from Eq. (9) and Eq. (10) that the LMA algorithm [10] is recoered when γ=0. 3. Performances of the NCLMA algorithm In this section the mean behaior and the second order moments of the proposed NCLMA algorithm are analyzed. We mae the following assumptions to analyze the NCLMA algorithm [9][1]: A1: is a wide-sense stationary sequence of i.i.d. Gaussian random ariables with zero-mean and positie definite autocorrelation matri R. A: is independent of the input process. A3: For any time instant α and w are statistically independent. A: For any time instant and wopt w are statistically independent. A5: enconditioned ( ) on is Gaussian with zero mean and e e. A6: he statistical dependence of and can be neglected. A7: and are statistically independent. Note that when parameters are chosen so that the steady-state ariance of α or w is e e 0

4 small then A can be approimately satisfied he computation compleity Unlie the LMA algorithm the parameters α and λ required to calculate in NCLMA algorithm. So computational compleity of the NCLMA algorithm is much than the LMA algorithm. Furthermore there is no sgn required to operate in NCLMF algorithm. But there is a higher order power operation for the error in NCLMF algorithm. For conenience the computational compleity of the NCLMA algorithm and that of other eisting algorithms is listed in able 1. As can be seen from able 1 the computational compleity of NCLMF algorithm and NCLMA algorithm is approimate. able 1. he computational compleity of NCLMA LMA and NCLMF algorithms Algorithm Multiplications Comparisons Additions ( ) ( ) 3 LMA N+ 1 N NCLM F N+9 0 N+5 1 NCLM A N+7 1 N Mean weight behaior model Based on Eq. (9) defining the weight error ector Epression Eq. (11) can be rewritten as e 1 sgn e e (1) e 1 e e 1 e 1 e 1 n e-n n0 0 has an initial alue of 0. So Substituting Eq. (1) into Eq. (10) hen defining we get 1 n n n (13) e (1) e (15) 1 n n n 1 n n0 Now substituting Eq. (16) into Eq. (1) n e (16)

5 1 1 1 n 3 1 e e sgn e n n0 1 1 n 3 sgn 1 e e en e sgn e n0 Redefining the time-arying term as 1 n n n (17) e (18) We obtain the final epression for the weight error ector update equation e e e e (19) 1 sgn sgn Using the aboe mentioned assumptions it can be shown that the mean behaior of the weight error ector is goerned by the following recursion e e e e E 1 E E sgn E sgn (0) Note that Eq. (0) is identical to the mean weight alue epression of the standard LMA algorithm ecept for the last term which can be written as follows E e sgn e n 3 1 n e n e e e e n0 n0 E 1 sgn E 1 sgn n 3 1 n E 1 e sgn 1 E sgn n e e e e n0 n0 1 We readily note that 1 following relation n0 1 n 3 n e (1) is statistically independent of resulting in the 1 1 n n 3 e n e e en e e n0 n0 E 1 sgn E 1 E sgn () Substituting Eq. (1) and Eq. () into Eq. (0) we get E 1 e e e e E E sgn E sgn n 3 1 n E E 1 1 E sgn e n e e n0 n0 he distribution function of Y e conditioned on F y p e y y y 1 e ep( e e is ) d e under A5 then the probability density function of Y e conditioned on is shown as Eq. (5) p y y F y 1 y ep( ) y e e F y (3) () (5)

6 he distribution function ofy e conditioned on is F y under A5 F y p e y y y 1 e ep( e e ) d e then the probability density function ofy e conditioned on So and and So hus E 1 E 1 n0 p y y F y y ep( ) e e is shown as Eq. (7) (6) (7) E e (8) e 3 3 Ee (9) e E e sgn( e ) E E e sgn( e ) e R E 1 n 3 1 Ee n 1 1 n0 1 n 3 en n 3 1 en n0 n n 3 1 n E E 1 1 E sgn en e e n0 n0 1 1 E 1 E 1 E n0 n0 1 n 3 1 n en e R 1 e R E 3 e n E n0 1 I 1 1 E R n0 3 e n e Condition for the stability of the mean weight error ector (as Eq. (3)) is gien by (30) (31) (3)

7 n0 3 e n e ma where ma is the maimum eigenalue associated with R. In addition when range of α is the LMA algorithm. As seen from Eq. (3) we will get 3.3. Second-order moment analysis he MSE is gien by E 0 0 as (33) this. E e E E tr R E (3) where E. Pre-multiplying Eq. (1) by its transpose form yields 1 sgn sgn sgn sgn e e e e e e e e e sgn e e sgn e e sgn e e sgn e e e e e aing the epected alue and we also assume that N is sufficiently large so that E N because of an essentially ergodic assumption such that the time aerage oer the taps is equal to the ensemble aerage. E 1 e e e e e e E E sgn E sgn E sgn E e sgn e E e E e E e e e E E E sgn E E E e E E E E E E e e o further simplify Eq. (36) we can tae adantage of Eq. (13) and Eq. (17) in [11] i.e. E E (37) (35) (36)

8 E e E E 6 E 6 E E 6 N E 6( ) E 3( ) E N N Substituting Eq. (37) and Eq. (38) into Eq. (36) E 1 E E E E E E e e E E E e So 6 E E E 6( ) E 3( ) E N N N 1 E e 6 E E E 6( ) 3( ) N N N Eq. (0) will epress as Eq. (1) when discard. 1 N E e 6( ) E E N E E E o further simplify Eq. (6) we can tae adantage of Eq. (31) Eq. () Eq. (3) and Eq. () i.e. and and 1 1n 1 1 n0 1 n0 6 E e e e ( 1 n ) 1 1 ( ) (38) (39) (0) (1) () 6 y ep( )d y (3) 0 e y

9 So 1 E 1 n0 1 n 3 1 E en 1 6 n0 1 n 3 1 E e E n en 1 1 n e n e n n A e n e n0 n0 1 1 A B ( N ) 1 A e n ( ) e n e n 0 n n n n0 n0 1 1 A B N 1 15 E A e n 0 0 ( ) e n e n 0 n n n n0 n0 where A 1 1 and B 1 1 (7) (8). As β should be chosen as a small alue to achiee a desired MSE and in addition according to the eisting literature [3][6] we can now and when0. t Eq. (8) will epress as Eq. (9) when discard for t. 1 A 1 B A e n e n0 n0 1 1 A B ( N ) 1 A e n ( ) e n e n 0 n n n n0 n0 1 1 A B N 1 15 E A e n 0 0 ( ) e n e n 0 n n n n0 n0 So where 1 A D( ) e A B F( ) ( ) 6( N ) 1 A D( ) A B F( ) N 1 ( ) E A D ( ) 1 3 en n C ( ) D( ) C( ) and F( ) 15 C( ). n 0 en n0 n0 herefore we obtain an epression of the behaior of the NCLMA algorithm i.e. where 1 e e (9) (50) f N g N (51)

10 f N e A B F( ) (5) 1 ( ) 6( ) 1 ( ) A D e N A D ( ) A B F( ) g N 1 ( ) E e N A D ( ) (53) he term f N e influences the conergence rate of the algorithm. It can be noticed that the fastest conergence mode is obtained when the function from Eq. (5) reaches its imum. e ( ) A D (5) A B F( ) 6( N ) 1 A D( ) ( ) For the NCLMA algorithm to be stable in the mean-square sense its behaior must decrease monotonically with an increase of. e ( ) 0 A D (55) A B F( ) ( N ) 1 A D( ) ( ) Assug that the NCLMA algorithm has conerged to the steady state 1 (56) he is obtained by substituting Eq. (56) into Eq. (51) as follows A( ) B( ) F( ) N N A( ) D 1 E A B F( ) 6( ) 1 A D e N A D ( ) Finally the MSE is gien by incorporating Eq. (57) into Eq. (3) A( ) B( ) F( ) N ( ) 1 E N A D MSE A B F( ) 6( ) 1 A D e N A D ( ) In addition we can now 0 or MSE when In the real practical world and are unnown howeer we can estimate e by using equation Eq. (59) [13][1]. 1 e = pp + (1 ) a 0. where p Ee is a small positie number to aoid that the denoator of Eq. (59) is infinite when 0. Although in Eq. (59) both p is also unnownwe can estimate this parameter by using Eq. (60). e (57) (58) and (59) p p (1 ) e (60)

11 where he alue ofe is listed in able. able. he alue of for different distributes E Gaussian[15] Uniform[15] Binary[15] Rayleigh[16] Eponential[17] E Finally the NCLMA algorithm consists of Eq.(9)-(11) Eq.(5) and Eq.(59)-(60).. Simulation results his section presents simulations in the contet of system identification with arious noises when the system is stationary or non-stationary to illustrate the accuracy of the NCLMA algorithm. he length of the unnown coefficient ector is ; the unnown coefficient is w opt. he uncorrelated input signal is a Gaussian white noise signal with ariance and the correlated input signal is calculated by y 0.8y. In all the eperiments the coefficient ectors are initialized as a zero 1 1 ector. 10 log10 opt w w is used to measure the performance of algorithms. Fig. 1 compares the analytical (from Eq. (57)) and eperimental results of the for the proposed NCLMA algorithm (β=0.001and γ=1000) for input with Uniform noise in stationary enironment with an SNR=0 db close agreement between theory and simulation is obtained as depicted in Fig. 1. he unnown coefficient is w opt =[ ]. he results are obtained by Monte Carlo simulations with 10 independent running. Iteration number is o understand clearly the improement brought about the NCLMA algorithm (β=0.001and γ=1000) when the input the time-arying step-size is depicted in Fig. for input with Uniform noise in stationary enironment with an SNR=0 db. he unnown coefficient is w opt =[ ]. he time-arying step-size of the NCLMA algorithm is shown in Fig.. Similarly the time ariation of of the NCLMA algorithm is depicted in Fig. 3. As obsered from this figure rises sharply and then decreases swiftly to reach zero. he results are obtained by Monte Carlo simulations with 10 independent running. Iteration number is Fig. displays the learning cures of the LMA (μ=0.01) NCLMA (β=0.001and γ=10000) and ZNCLMA (β=0.001and γ=1000) algorithms when the noise statistic is Eponential distributed signal with for an SNR=10dB with y input in nonstationary enironment. he results are obtained by Monte Carlo simulations with 30 independent running. Iteration number is he unnown coefficient is w w 1 υ where y w L opt opt w opt =[ ] υ is an i.i.d. Gaussian sequence with m 0 and Here too the ecellent performance of the proposed NCLMA algorithm is maintained and therefore a consistency in performance is achieed by the proposed NCLMA algorithm. Net the choice of β and γ can also affect the performance of the NCLMA

12 algorithm. Results in Fig. 5 for different alues of β and γ show clearly that een in etreme cases. he system noise is Uniform distributed oer the interal (-3 3). he unnown coefficient is w opt =[ ] with input. he results are obtained by Monte Carlo simulations with 10 independent running. Iteration number is he cures of the NCLMA algorithm (β=0.001and γ=5000) with SNR=0dB. he performance of the NCLMA algorithm did not deteriorate much. he best performance was achieed by increasing β to 0.01 and decreasing γ to 1000 for a constant steady-state misadjustment obtained by the NCLMA algorithm. Fig.6 summarizes the performance of the proposed NCLMA algorithm (β=0.001 and γ=1000) in the fourth different noise enironments with an SNR of 0dB with y input in nonstationary enironment. he results are obtained by Monte Carlo simulations with 30 independent running. Iteration number is he unnown coefficient is wopt wopt 1 υ where w opt =[ ] is an i.i.d. Gaussian sequence with m 0 and From this figure that the best performance is obtained when the noise statistic is Rayleigh while the worst performance is obtained when the noise statistic is Uniform. Moreoer to further test its performance Fig.7 display the learning cures of the NCLMF (α=0.001 β= and γ=500) and NCLMA algorithms (β=0.001and γ=10000) with SNR= 0dB when the noise statistic is Gaussian white noise with υ m 0 and 1. he unnown coefficient is w opt =[ ]. he correlated input signal was gien by y 0.8y 1. Fig.8 display the learning cures of the NCLMA algorithms (β=0.001and γ=0000) and ZNCLMA (β= and γ=000) algorithms when the noise statistics are Eponential distributed signal with. he unnown coefficient is w w 1 υ where w opt =[ ] υ opt opt is an i.i.d. Gaussian sequence with m 0 and he correlated input signal was gien by y 0.8y 1. he results are obtained by Monte Carlo simulations with 10 independent running. Iteration number is he unnown coefficient is w opt =[ ]. he correlated input signal was gien by y 0.8y 1. Here too the ecellent performance of the proposed NCLMA (β=0.001 and γ=000) algorithm is maintained and therefore a consistency in performance is achieed by the proposed NCLMA algorithm. Moreoer to further test its performance the noise statistics are changed to Rayleigh distributed signal with for an SNR=0 db and as depicted from Fig. 9 a similar behaior as the former ones is obtained by the proposed NCLMA algorithm in this enironment. he results are obtained by Monte Carlo simulations with 10 independent running. Iteration number is he tuning parameters for this part of the ZNCLMA algorithm are β=0.003 and γ=100.

13 0 10 NCLMA-simulation NCLMA-theory Fig.1. learning cure of the NCLMA algorithm NCLMA ime arying step-size Fig.. Behaior of time-arying step-size of NCLMA algorithm NCLMA 0.0 Behaior of λ Fig.3. Behaior of of NCLMA algorithm

14 LMA NCLMA ZNCLMA Fig.. Learning cures for the LMA NCLMA and ZNCLMA algorithms NCLMA β= γ=0000 NCLMA β= γ=10000 NCLMA β=0.001 γ=5000 NCLMA β=0.01 γ= Fig.5. he effect of β and γ on the conergence behaior of the NCLMA algorithm NCLMA->Uniform noise NCLMA->Epential noise NCLMA->Binary noise NCLMA->Rayleigh noise Fig.6. Learning cures for the NCLMA algorithm in the fourth different noise enironments

15 NCLMF NCLMA Fig.7. learning cures of the NCLMF and NCLMA algorithms NCLMA ZNCLMA Fig.8. Behaior of time-ary step-size of NCLMA and ZNCLMA algorithms NCLMA ZNCLMA Fig.9. learning cures of the NCLMF and NCLMA algorithms.

16 5. Conclusion In this paper we proposed a noise constrained LMA-type algorithm (NCLMA) for FIR channel estimation and studied its performance both analytically and simulations. he NCLMA algorithm is obtained by constraining the cost function of the standard LMA algorithm to the third-order moment of the additie noise. he main aim of this wor is to derie the NCLMA algorithm analyze optimized α and assess the performance in steady-state in different noise enironments when the system is stationary or non-stationary. Compare with the NCLMF and LMA algorithms the NCLMA enjoys a superior performance. he techniques introduced here may be applicable to plant identification. Acnowledgments his wor was supported by National Natural Science Foundation of China ( ) and the State Key Laboratory of Intelligent Control and Decision of Comple Systems of Beijing Institute of echnology. References [1] H. Sayed. Adaptie Filters [M]. Hoboen N USA: ohn Wiley & Sons Inc [] P.S.R. Diniz Adaptie Filtering (Fourth Edi.) [M]. Springer US 013 [3] Wei Y Gelfand S B Krogmeier V. Noise-constrained least mean squares algorithm []. IEEE rans. Signal Process. 9(9) (001) [] R. Kwong and E. W. ohnston. A ariable step size LMS algorithm []. IEEE rans. Signal Process. 0(7) (199) [5] Zhao H Yu Y Gao S Zeng X He Z. A new normalized LMA algorithm and its performance analysis []. Signal Process. 105(1) (01) [6] Zerguine A Moinuddin M Imam S A A. A noise constrained least mean fourth (NCLMF) adaptie algorithm []. Signal Process. 91(1) (011) [7] Eweda E. Dependence of the Stability of the Least-mean Fourth Algorithm on arget Weights Non-Stationarity []. IEEE rans. Signal Process. 6(7) (01) [8] Lee Y H in D M Sang D K Cho S H. Performance of least-mean absolute third (LMA) adaptie algorithm in arious noise enironments []. Electron. Lett. 3(3) (1998) 1-3. [9] Guan S Li Z. Nonparametric ariable step-size LMA algorithm []. Circ. Syst. & Signal Process. 016:1-18. [10] Cho S Sang D K. Adaptie filters based on the high order error statistics[c]// Circuits and Systems IEEE Asia Pacific Conference on (1996) [11] Hubscher P I Bermudez C M Nascimento V H. A Mean-Square Stability Analysis of the Least Mean Fourth Adaptie Algorithm []. IEEE rans. Signal Process. 55(8) (007) [1] Cho S H Mathews V. racing analysis of the sign algorithm in nonstationary enironments []. IEEE rans. Acoustics Speech & Signal Processing 38(1) (1990)

17 [13] H.-C. Huang. Lee. A new ariable step-size NLMS algorithm and its performance analysis. IEEE rans. Signal Process. 60() (01) [1] Han-Sol Lee Seong-Eun Kim ae-woo Lee and Woo-in Song. A Variable Step- Size Diffusion LMS Algorithm for Distributed Estimation []. IEEE rans. Signal Process. 63(7) (015) [15] Eweda E Bershad N. Stochastic Analysis of a Stable Normalized Least Mean Fourth Algorithm for Adaptie Noise Canceling With a White Gaussian Reference []. IEEE rans. Signal Process. 60(1) (01) [16] K. Hirano Rayleigh Distribution (Wiley London 01) [17] Spiegel M R. Mathematical handboo of formulas and tables [M]. McGraw-Hill 01.

Notes on Linear Minimum Mean Square Error Estimators

Notes on Linear Minimum Mean Square Error Estimators Notes on Linear Minimum Mean Square Error Estimators Ça gatay Candan January, 0 Abstract Some connections between linear minimum mean square error estimators, maximum output SNR filters and the least square

More information

Sample Average Approximation for Stochastic Empty Container Repositioning

Sample Average Approximation for Stochastic Empty Container Repositioning Sample Aerage Approimation for Stochastic Empty Container Repositioning E Peng CHEW a, Loo Hay LEE b, Yin LOG c Department of Industrial & Systems Engineering ational Uniersity of Singapore Singapore 9260

More information

Partial Update LMS Algorithms

Partial Update LMS Algorithms JOURNAL OF L A TEX CLASS FILES, VOL. 1, NO. 11, NOVEMBER 00 1 Partial Update LMS Algorithms Mahesh Godaarti, Member, IEEE, and Alfred Hero, Fellow, IEEE Abstract Partial updating of LMS filter coefficients

More information

Trajectory Estimation for Tactical Ballistic Missiles in Terminal Phase Using On-line Input Estimator

Trajectory Estimation for Tactical Ballistic Missiles in Terminal Phase Using On-line Input Estimator Proc. Natl. Sci. Counc. ROC(A) Vol. 23, No. 5, 1999. pp. 644-653 Trajectory Estimation for Tactical Ballistic Missiles in Terminal Phase Using On-line Input Estimator SOU-CHEN LEE, YU-CHAO HUANG, AND CHENG-YU

More information

Probabilistic Engineering Design

Probabilistic Engineering Design Probabilistic Engineering Design Chapter One Introduction 1 Introduction This chapter discusses the basics of probabilistic engineering design. Its tutorial-style is designed to allow the reader to quickly

More information

arxiv: v1 [physics.comp-ph] 17 Jan 2014

arxiv: v1 [physics.comp-ph] 17 Jan 2014 An efficient method for soling a correlated multi-item inentory system Chang-Yong Lee and Dongu Lee The Department of Industrial & Systems Engineering, Kongu National Uniersity, Kongu 34-70 South Korea

More information

Adaptive Filter Theory

Adaptive Filter Theory 0 Adaptive Filter heory Sung Ho Cho Hanyang University Seoul, Korea (Office) +8--0-0390 (Mobile) +8-10-541-5178 dragon@hanyang.ac.kr able of Contents 1 Wiener Filters Gradient Search by Steepest Descent

More information

On the Stability of the Least-Mean Fourth (LMF) Algorithm

On the Stability of the Least-Mean Fourth (LMF) Algorithm XXI SIMPÓSIO BRASILEIRO DE TELECOMUNICACÕES-SBT 4, 6-9 DE SETEMBRO DE 4, BELÉM, PA On the Stability of the Least-Mean Fourth (LMF) Algorithm Vítor H. Nascimento and José Carlos M. Bermudez + Abstract We

More information

Error Vector Normalized Adaptive Algorithm Applied to Adaptive Noise Canceller and System Identification

Error Vector Normalized Adaptive Algorithm Applied to Adaptive Noise Canceller and System Identification American J. of Engineering and Applied Sciences 3 (4): 710-717, 010 ISSN 1941-700 010 Science Publications Error Vector Normalized Adaptive Algorithm Applied to Adaptive Noise Canceller and System Identification

More information

SUPPLEMENTARY MATERIAL. Authors: Alan A. Stocker (1) and Eero P. Simoncelli (2)

SUPPLEMENTARY MATERIAL. Authors: Alan A. Stocker (1) and Eero P. Simoncelli (2) SUPPLEMENTARY MATERIAL Authors: Alan A. Stocker () and Eero P. Simoncelli () Affiliations: () Dept. of Psychology, Uniersity of Pennsylania 34 Walnut Street 33C Philadelphia, PA 94-68 U.S.A. () Howard

More information

Mean-variance receding horizon control for discrete time linear stochastic systems

Mean-variance receding horizon control for discrete time linear stochastic systems Proceedings o the 17th World Congress The International Federation o Automatic Control Seoul, Korea, July 6 11, 008 Mean-ariance receding horizon control or discrete time linear stochastic systems Mar

More information

Review of Matrices and Vectors 1/45

Review of Matrices and Vectors 1/45 Reiew of Matrices and Vectors /45 /45 Definition of Vector: A collection of comple or real numbers, generally put in a column [ ] T "! Transpose + + + b a b a b b a a " " " b a b a Definition of Vector

More information

0 a 3 a 2 a 3 0 a 1 a 2 a 1 0

0 a 3 a 2 a 3 0 a 1 a 2 a 1 0 Chapter Flow kinematics Vector and tensor formulae This introductory section presents a brief account of different definitions of ector and tensor analysis that will be used in the following chapters.

More information

Experimental link of coarsening rate and volume distribution in dry foam

Experimental link of coarsening rate and volume distribution in dry foam August 2012 EPL, 99 (2012 48003 doi:.1209/0295-5075/99/48003 www.epljournal.org Eperimental link of coarsening rate and olume distribution in dry foam J. Lambert 1, F. Graner 2,3, R. Delannay 1 and I.

More information

Astrometric Errors Correlated Strongly Across Multiple SIRTF Images

Astrometric Errors Correlated Strongly Across Multiple SIRTF Images Astrometric Errors Correlated Strongly Across Multiple SIRTF Images John Fowler 28 March 23 The possibility exists that after pointing transfer has been performed for each BCD (i.e. a calibrated image

More information

Optimized Concatenated LDPC Codes for Joint Source-Channel Coding

Optimized Concatenated LDPC Codes for Joint Source-Channel Coding Optimized Concatenated LDPC Codes for Joint Source-Channel Coding Maria Fresia, Fernando Pérez-Cruz, H. Vincent Poor Department of Electrical Engineering, Princeton Uniersity, Princeton, New Jersey 08544

More information

Evolution Analysis of Iterative LMMSE-APP Detection for Coded Linear System with Cyclic Prefixes

Evolution Analysis of Iterative LMMSE-APP Detection for Coded Linear System with Cyclic Prefixes Eolution Analysis of Iteratie LMMSE-APP Detection for Coded Linear System with Cyclic Prefixes Xiaoun Yuan Qinghua Guo and Li Ping Member IEEE Department of Electronic Engineering City Uniersity of Hong

More information

NUMERICAL COMPUTATION OF THE CAPACITY OF CONTINUOUS MEMORYLESS CHANNELS

NUMERICAL COMPUTATION OF THE CAPACITY OF CONTINUOUS MEMORYLESS CHANNELS NUMERICAL COMPUTATION OF THE CAPACITY OF CONTINUOUS MEMORYLESS CHANNELS Justin Dauwels Dept. of Information Technology and Electrical Engineering ETH, CH-8092 Zürich, Switzerland dauwels@isi.ee.ethz.ch

More information

IS NEGATIVE STEP SIZE LMS ALGORITHM STABLE OPERATION POSSIBLE?

IS NEGATIVE STEP SIZE LMS ALGORITHM STABLE OPERATION POSSIBLE? IS NEGATIVE STEP SIZE LMS ALGORITHM STABLE OPERATION POSSIBLE? Dariusz Bismor Institute of Automatic Control, Silesian University of Technology, ul. Akademicka 16, 44-100 Gliwice, Poland, e-mail: Dariusz.Bismor@polsl.pl

More information

Optimal Joint Detection and Estimation in Linear Models

Optimal Joint Detection and Estimation in Linear Models Optimal Joint Detection and Estimation in Linear Models Jianshu Chen, Yue Zhao, Andrea Goldsmith, and H. Vincent Poor Abstract The problem of optimal joint detection and estimation in linear models with

More information

Section 6: PRISMATIC BEAMS. Beam Theory

Section 6: PRISMATIC BEAMS. Beam Theory Beam Theory There are two types of beam theory aailable to craft beam element formulations from. They are Bernoulli-Euler beam theory Timoshenko beam theory One learns the details of Bernoulli-Euler beam

More information

Parameters Identification of Equivalent Circuit Diagrams for Li-Ion Batteries

Parameters Identification of Equivalent Circuit Diagrams for Li-Ion Batteries Parameters Identification of Equialent Circuit Diagrams for Li-Ion eries Ahmad ahmoun, Helmuth Biechl Uniersity of Applied ciences Kempten Ahmad.ahmoun@stud.fh-empten.de, biechl@fh-empten.de Abstract-eries

More information

A low intricacy variable step-size partial update adaptive algorithm for Acoustic Echo Cancellation USNRao

A low intricacy variable step-size partial update adaptive algorithm for Acoustic Echo Cancellation USNRao ISSN: 77-3754 International Journal of Engineering and Innovative echnology (IJEI Volume 1, Issue, February 1 A low intricacy variable step-size partial update adaptive algorithm for Acoustic Echo Cancellation

More information

Econometrics II - EXAM Outline Solutions All questions have 25pts Answer each question in separate sheets

Econometrics II - EXAM Outline Solutions All questions have 25pts Answer each question in separate sheets Econometrics II - EXAM Outline Solutions All questions hae 5pts Answer each question in separate sheets. Consider the two linear simultaneous equations G with two exogeneous ariables K, y γ + y γ + x δ

More information

Reduced-cost combination of adaptive filters for acoustic echo cancellation

Reduced-cost combination of adaptive filters for acoustic echo cancellation Reduced-cost combination of adaptive filters for acoustic echo cancellation Luis A. Azpicueta-Ruiz and Jerónimo Arenas-García Dept. Signal Theory and Communications, Universidad Carlos III de Madrid Leganés,

More information

A. Idesman. Keywords: time integration, spurious oscillations, numerical dispersion

A. Idesman. Keywords: time integration, spurious oscillations, numerical dispersion COMPDYN 0 rd ECCOMAS Thematic Conference on Computational Methods in Structural Dynamics and Earthquake Engineering M. Papadrakakis, M. Fragiadakis, V. Pleris (eds.) Corfu, Greece, -8 May 0 ACCURATE NUMERICAL

More information

VISUAL PHYSICS ONLINE RECTLINEAR MOTION: UNIFORM ACCELERATION

VISUAL PHYSICS ONLINE RECTLINEAR MOTION: UNIFORM ACCELERATION VISUAL PHYSICS ONLINE RECTLINEAR MOTION: UNIFORM ACCELERATION Predict Obsere Explain Exercise 1 Take an A4 sheet of paper and a heay object (cricket ball, basketball, brick, book, etc). Predict what will

More information

Сollisionless damping of electron waves in non-maxwellian plasma 1

Сollisionless damping of electron waves in non-maxwellian plasma 1 http:/arxi.org/physics/78.748 Сollisionless damping of electron waes in non-mawellian plasma V. N. Soshnio Plasma Physics Dept., All-Russian Institute of Scientific and Technical Information of the Russian

More information

Variable Learning Rate LMS Based Linear Adaptive Inverse Control *

Variable Learning Rate LMS Based Linear Adaptive Inverse Control * ISSN 746-7659, England, UK Journal of Information and Computing Science Vol., No. 3, 6, pp. 39-48 Variable Learning Rate LMS Based Linear Adaptive Inverse Control * Shuying ie, Chengjin Zhang School of

More information

Assignment 4 (Solutions) NPTEL MOOC (Bayesian/ MMSE Estimation for MIMO/OFDM Wireless Communications)

Assignment 4 (Solutions) NPTEL MOOC (Bayesian/ MMSE Estimation for MIMO/OFDM Wireless Communications) Assignment 4 Solutions NPTEL MOOC Bayesian/ MMSE Estimation for MIMO/OFDM Wireless Communications The system model can be written as, y hx + The MSE of the MMSE estimate ĥ of the aboe mentioned system

More information

STOCHASTIC INFORMATION GRADIENT ALGORITHM BASED ON MAXIMUM ENTROPY DENSITY ESTIMATION. Badong Chen, Yu Zhu, Jinchun Hu and Ming Zhang

STOCHASTIC INFORMATION GRADIENT ALGORITHM BASED ON MAXIMUM ENTROPY DENSITY ESTIMATION. Badong Chen, Yu Zhu, Jinchun Hu and Ming Zhang ICIC Express Letters ICIC International c 2009 ISSN 1881-803X Volume 3, Number 3, September 2009 pp. 1 6 STOCHASTIC INFORMATION GRADIENT ALGORITHM BASED ON MAXIMUM ENTROPY DENSITY ESTIMATION Badong Chen,

More information

Adaptive Filtering Part II

Adaptive Filtering Part II Adaptive Filtering Part II In previous Lecture we saw that: Setting the gradient of cost function equal to zero, we obtain the optimum values of filter coefficients: (Wiener-Hopf equation) Adaptive Filtering,

More information

Convergence Evaluation of a Random Step-Size NLMS Adaptive Algorithm in System Identification and Channel Equalization

Convergence Evaluation of a Random Step-Size NLMS Adaptive Algorithm in System Identification and Channel Equalization Convergence Evaluation of a Random Step-Size NLMS Adaptive Algorithm in System Identification and Channel Equalization 1 Shihab Jimaa Khalifa University of Science, Technology and Research (KUSTAR) Faculty

More information

SLIP MODEL PERFORMANCE FOR MICRO-SCALE GAS FLOWS

SLIP MODEL PERFORMANCE FOR MICRO-SCALE GAS FLOWS 3th AIAA Thermophysics Conference 3- June 3, Orlando, Florida AIAA 3-5 SLIP MODEL PERFORMANCE FOR MICRO-SCALE GAS FLOWS Matthew J. McNenly* Department of Aerospace Engineering Uniersity of Michigan, Ann

More information

Network Flow Problems Luis Goddyn, Math 408

Network Flow Problems Luis Goddyn, Math 408 Network Flow Problems Luis Goddyn, Math 48 Let D = (V, A) be a directed graph, and let s, t V (D). For S V we write δ + (S) = {u A : u S, S} and δ (S) = {u A : u S, S} for the in-arcs and out-arcs of S

More information

Dynamic Vehicle Routing with Moving Demands Part II: High speed demands or low arrival rates

Dynamic Vehicle Routing with Moving Demands Part II: High speed demands or low arrival rates ACC 9, Submitted St. Louis, MO Dynamic Vehicle Routing with Moing Demands Part II: High speed demands or low arrial rates Stephen L. Smith Shaunak D. Bopardikar Francesco Bullo João P. Hespanha Abstract

More information

Alpha-Stable Distributions in Signal Processing of Audio Signals

Alpha-Stable Distributions in Signal Processing of Audio Signals Alpha-Stable Distributions in Signal Processing of Audio Signals Preben Kidmose, Department of Mathematical Modelling, Section for Digital Signal Processing, Technical University of Denmark, Building 3,

More information

Asymptotic Normality of an Entropy Estimator with Exponentially Decaying Bias

Asymptotic Normality of an Entropy Estimator with Exponentially Decaying Bias Asymptotic Normality of an Entropy Estimator with Exponentially Decaying Bias Zhiyi Zhang Department of Mathematics and Statistics Uniersity of North Carolina at Charlotte Charlotte, NC 28223 Abstract

More information

The hierarchical real-time control of high speed trains for automatic train operation

The hierarchical real-time control of high speed trains for automatic train operation Computers in Railways XIV 17 The hierarchical real-time control of high speed trains for automatic train operation Q. Y. Wang, P. Wu, Z. C. Liang & X. Y. Feng Southwest Jiaotong Uniersity, China Abstract

More information

arxiv: v1 [stat.ml] 15 Feb 2018

arxiv: v1 [stat.ml] 15 Feb 2018 1 : A New Algorithm for Streaming PCA arxi:1802.054471 [stat.ml] 15 Feb 2018 Puyudi Yang, Cho-Jui Hsieh, Jane-Ling Wang Uniersity of California, Dais pydyang, chohsieh, janelwang@ucdais.edu Abstract In

More information

Tools for Investigation of Dynamics of DC-DC Converters within Matlab/Simulink

Tools for Investigation of Dynamics of DC-DC Converters within Matlab/Simulink Tools for Inestigation of Dynamics of DD onerters within Matlab/Simulink Riga Technical Uniersity, Riga, Latia Email: pikulin03@inbox.l Dmitry Pikulin Abstract: In this paper the study of complex phenomenon

More information

A Guidance Law for a Mobile Robot for Coverage Applications: A Limited Information Approach

A Guidance Law for a Mobile Robot for Coverage Applications: A Limited Information Approach Proceedings of the Third International Conference on Adances in Control and Optimization of Dynamical Systems, Indian Institute of Technology anpur, India, March 1-15, 014 Frcsh1.6 A Guidance Law for a

More information

An Elastic Contact Problem with Normal Compliance and Memory Term

An Elastic Contact Problem with Normal Compliance and Memory Term Machine Dynamics Research 2012, Vol. 36, No 1, 15 25 Abstract An Elastic Contact Problem with Normal Compliance and Memory Term Mikäel Barboteu, Ahmad Ramadan, Mircea Sofonea Uniersité de Perpignan, France

More information

Estimation of Efficiency with the Stochastic Frontier Cost. Function and Heteroscedasticity: A Monte Carlo Study

Estimation of Efficiency with the Stochastic Frontier Cost. Function and Heteroscedasticity: A Monte Carlo Study Estimation of Efficiency ith the Stochastic Frontier Cost Function and Heteroscedasticity: A Monte Carlo Study By Taeyoon Kim Graduate Student Oklahoma State Uniersity Department of Agricultural Economics

More information

Chapter 8: MULTIPLE CONTINUOUS RANDOM VARIABLES

Chapter 8: MULTIPLE CONTINUOUS RANDOM VARIABLES Charles Boncelet Probabilit Statistics and Random Signals" Oord Uniersit Press 06. ISBN: 978-0-9-0005-0 Chapter 8: MULTIPLE CONTINUOUS RANDOM VARIABLES Sections 8. Joint Densities and Distribution unctions

More information

An Explicit Lower Bound of 5n o(n) for Boolean Circuits

An Explicit Lower Bound of 5n o(n) for Boolean Circuits An Eplicit Lower Bound of 5n o(n) for Boolean Circuits Kazuo Iwama, Oded Lachish, Hiroki Morizumi, and Ran Raz Graduate School of Informatics, Kyoto Uniersity, Kyoto, JAPAN {iwama, morizumi}@kuis.kyoto-u.ac.jp

More information

Predictive Coding. Lossy or lossless. Feedforward or feedback. Intraframe or interframe. Fixed or Adaptive

Predictive Coding. Lossy or lossless. Feedforward or feedback. Intraframe or interframe. Fixed or Adaptive Predictie Coding Predictie coding is a compression tecnique based on te difference between te original and predicted alues. It is also called DPCM Differential Pulse Code Modulation Lossy or lossless Feedforward

More information

Non-Surjective Finite Alphabet Iterative Decoders

Non-Surjective Finite Alphabet Iterative Decoders IEEE ICC 2016 - Communications Theory Non-Surjectie Finite Alphabet Iteratie Decoders Thien Truong Nguyen-Ly, Khoa Le,V.Sain, D. Declercq, F. Ghaffari and O. Boncalo CEA-LETI, MINATEC Campus, Grenoble,

More information

Mathisson s New Mechanics : Its Aims and Realisation. W G Dixon, Churchill College, Cambridge, England

Mathisson s New Mechanics : Its Aims and Realisation. W G Dixon, Churchill College, Cambridge, England Lecture : ims Mathisson s New Mechanics : Its ims Realisation W G Dixon, Churchill College, Cambridge, Engl It gies me great pleasure to be inited to speak at this meeting on the life work of Myron Mathisson

More information

Cramér-Rao Bounds for Estimation of Linear System Noise Covariances

Cramér-Rao Bounds for Estimation of Linear System Noise Covariances Journal of Mechanical Engineering and Automation (): 6- DOI: 593/jjmea Cramér-Rao Bounds for Estimation of Linear System oise Covariances Peter Matiso * Vladimír Havlena Czech echnical University in Prague

More information

Sparse Functional Regression

Sparse Functional Regression Sparse Functional Regression Junier B. Olia, Barnabás Póczos, Aarti Singh, Jeff Schneider, Timothy Verstynen Machine Learning Department Robotics Institute Psychology Department Carnegie Mellon Uniersity

More information

v v Downloaded 01/11/16 to Redistribution subject to SEG license or copyright; see Terms of Use at

v v Downloaded 01/11/16 to Redistribution subject to SEG license or copyright; see Terms of Use at The pseudo-analytical method: application of pseudo-laplacians to acoustic and acoustic anisotropic wae propagation John T. Etgen* and Serre Brandsberg-Dahl Summary We generalize the pseudo-spectral method

More information

Nonlinear Disturbance Decoupling for a Mobile Robotic Manipulator over Uneven Terrain

Nonlinear Disturbance Decoupling for a Mobile Robotic Manipulator over Uneven Terrain Nonlinear Disturbance Decoupling for a Mobile Robotic Manipulator oer Uneen Terrain Joel Jimenez-Lozano Bill Goodwine Department of Aerospace and Mechanical Engineering Uniersity of Noe Dame Noe Dame IN

More information

An example of Lagrangian for a non-holonomic system

An example of Lagrangian for a non-holonomic system Uniersit of North Georgia Nighthaks Open Institutional Repositor Facult Publications Department of Mathematics 9-9-05 An eample of Lagrangian for a non-holonomic sstem Piotr W. Hebda Uniersit of North

More information

Signals and Spectra - Review

Signals and Spectra - Review Signals and Spectra - Review SIGNALS DETERMINISTIC No uncertainty w.r.t. the value of a signal at any time Modeled by mathematical epressions RANDOM some degree of uncertainty before the signal occurs

More information

DEVIL PHYSICS THE BADDEST CLASS ON CAMPUS AP PHYSICS

DEVIL PHYSICS THE BADDEST CLASS ON CAMPUS AP PHYSICS DEVIL PHYSICS THE BADDEST CLASS ON CAMPUS AP PHYSICS LSN 7-4: CONSERVATION OF ENERGY AND MOMENTUM IN COLLISIONS LSN 7-5: ELASTIC COLLISIONS IN ONE DIMENSION LSN 7-6: INELASTIC COLLISIONS Questions From

More information

Admission Control and Profitability Analysis in Dynamic Spectrum Access Data Networks

Admission Control and Profitability Analysis in Dynamic Spectrum Access Data Networks Admission Control and Profitability Analysis in Dynamic Spectrum Access Data Networks ABSTRACT Sinem Kockan Department of Electrical and Computer Engineering Boston Uniersity Boston, MA 02215 kockan@bu.edu

More information

Fractional-RWHet: An enhanced solver for solute transport with both spatiotemporal memory and conditioning on local aquifer properties

Fractional-RWHet: An enhanced solver for solute transport with both spatiotemporal memory and conditioning on local aquifer properties Fractional-RWHet: An enhanced soler for solute transport with both spatiotemporal memory and conditioning on local aquifer properties Zhang Yong 1, Daid A. Benson 2, Eric M. LaBolle 3, Donald M. Reees

More information

DESIGN METHOD BASED ON THE CONCEPT OF EMERGENCE AND ITS APPLICATION

DESIGN METHOD BASED ON THE CONCEPT OF EMERGENCE AND ITS APPLICATION DESIGN METHOD BASED ON THE CONCEPT OF EMERGENCE AND ITS APPLICATION Koichiro Sato¹ and Yoshiyuki Matsuoka 2 ¹ Graduate School of Science and Technology, Keio Uniersity, Yokohama, Japan Koichiro_Sato@a7.keio.jp,

More information

Dynamics ( 동역학 ) Ch.2 Motion of Translating Bodies (2.1 & 2.2)

Dynamics ( 동역학 ) Ch.2 Motion of Translating Bodies (2.1 & 2.2) Dynamics ( 동역학 ) Ch. Motion of Translating Bodies (. &.) Motion of Translating Bodies This chapter is usually referred to as Kinematics of Particles. Particles: In dynamics, a particle is a body without

More information

Submitted to Electronics Letters. Indexing terms: Signal Processing, Adaptive Filters. The Combined LMS/F Algorithm Shao-Jen Lim and John G. Harris Co

Submitted to Electronics Letters. Indexing terms: Signal Processing, Adaptive Filters. The Combined LMS/F Algorithm Shao-Jen Lim and John G. Harris Co Submitted to Electronics Letters. Indexing terms: Signal Processing, Adaptive Filters. The Combined LMS/F Algorithm Shao-Jen Lim and John G. Harris Computational Neuro-Engineering Laboratory University

More information

A nonparametric confidence interval for At-Risk-of-Poverty-Rate: an example of application

A nonparametric confidence interval for At-Risk-of-Poverty-Rate: an example of application A nonparametric confidence interal for At-Ris-of-Poerty-Rate: an example of application Wojciech Zielińsi Department of Econometrics and Statistics Warsaw Uniersity of Life Sciences Nowoursynowsa 159,

More information

STEADY-STATE MEAN SQUARE PERFORMANCE OF A SPARSIFIED KERNEL LEAST MEAN SQUARE ALGORITHM.

STEADY-STATE MEAN SQUARE PERFORMANCE OF A SPARSIFIED KERNEL LEAST MEAN SQUARE ALGORITHM. STEADY-STATE MEAN SQUARE PERFORMANCE OF A SPARSIFIED KERNEL LEAST MEAN SQUARE ALGORITHM Badong Chen 1, Zhengda Qin 1, Lei Sun 2 1 Institute of Artificial Intelligence and Robotics, Xi an Jiaotong University,

More information

Performance Analysis and Enhancements of Adaptive Algorithms and Their Applications

Performance Analysis and Enhancements of Adaptive Algorithms and Their Applications Performance Analysis and Enhancements of Adaptive Algorithms and Their Applications SHENGKUI ZHAO School of Computer Engineering A thesis submitted to the Nanyang Technological University in partial fulfillment

More information

ME224 Lab 5 - Thermal Diffusion

ME224 Lab 5 - Thermal Diffusion ME4 Lab 5 ME4 Lab 5 - hermal Diffusion (his lab is adapted from IBM-PC in the laboratory by B G homson & A F Kuckes, Chapter 5) 1. Introduction he experiments which you will be called upon to do in this

More information

Towards Green Distributed Storage Systems

Towards Green Distributed Storage Systems Towards Green Distributed Storage Systems Abdelrahman M. Ibrahim, Ahmed A. Zewail, and Aylin Yener Wireless Communications and Networking Laboratory (WCAN) Electrical Engineering Department The Pennsylania

More information

متلب سایت MatlabSite.com

متلب سایت MatlabSite.com Vehicle Velocity Estimation based on Data fusion by Kalman Filtering for ABS Melika Amiri Graduate student in electrical engineering, Islamic Azad uniersity, Science and Research ehran, Iran Melika_64@yahoo.com

More information

KNOWN approaches for improving the performance of

KNOWN approaches for improving the performance of IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS II: EXPRESS BRIEFS, VOL. 58, NO. 8, AUGUST 2011 537 Robust Quasi-Newton Adaptive Filtering Algorithms Md. Zulfiquar Ali Bhotto, Student Member, IEEE, and Andreas

More information

Progress In Electromagnetics Research Symposium 2005, Hangzhou, China, August

Progress In Electromagnetics Research Symposium 2005, Hangzhou, China, August Progress In Electromagnetics Research Symposium 2005, Hangzhou, China, August 22-26 115 A Noel Technique for Localizing the Scatterer in Inerse Profiling of Two Dimensional Circularly Symmetric Dielectric

More information

Target Trajectory Estimation within a Sensor Network

Target Trajectory Estimation within a Sensor Network Target Trajectory Estimation within a Sensor Network Adrien Ickowicz IRISA/CNRS, 354, Rennes, J-Pierre Le Cadre, IRISA/CNRS,354, Rennes,France Abstract This paper deals with the estimation of the trajectory

More information

Solution to 1-D consolidation of non-homogeneous soft clay *

Solution to 1-D consolidation of non-homogeneous soft clay * Xie et al. / J Zhejiang Uni SCI 25 6A(Suppl. I):29-34 29 Journal of Zhejiang Uniersity SCIENCE ISSN 19-395 http://www.zju.edu.cn/jzus E-mail: jzus@zju.edu.cn Solution to 1-D consolidation of non-homogeneous

More information

ERAD THE SEVENTH EUROPEAN CONFERENCE ON RADAR IN METEOROLOGY AND HYDROLOGY

ERAD THE SEVENTH EUROPEAN CONFERENCE ON RADAR IN METEOROLOGY AND HYDROLOGY Multi-beam raindrop size distribution retrieals on the oppler spectra Christine Unal Geoscience and Remote Sensing, TU-elft Climate Institute, Steinweg 1, 68 CN elft, Netherlands, c.m.h.unal@tudelft.nl

More information

EXPERIMENT 8 BALLISTIC PENDULUM. Figure 1 Setup to determine the initial speed of the projectile using the Blackwood Pendulum

EXPERIMENT 8 BALLISTIC PENDULUM. Figure 1 Setup to determine the initial speed of the projectile using the Blackwood Pendulum EXPERIMENT 8 BALLISTIC PENDULUM I. Introduction. The objectie of this eperiment is to determine the initial elocity of a projectile fired from a gun by two methods. In the first the projectile undergoes

More information

Chapter 16. Kinetic Theory of Gases. Summary. molecular interpretation of the pressure and pv = nrt

Chapter 16. Kinetic Theory of Gases. Summary. molecular interpretation of the pressure and pv = nrt Chapter 16. Kinetic Theory of Gases Summary molecular interpretation of the pressure and pv = nrt the importance of molecular motions elocities and speeds of gas molecules distribution functions for molecular

More information

NEW FRONTIERS IN APPLIED PROBABILITY

NEW FRONTIERS IN APPLIED PROBABILITY J. Appl. Prob. Spec. Vol. 48A, 183 194 (2011) Applied Probability Trust 2011 NEW FRONTIERS IN APPLIED PROBABILITY A Festschrift for SØREN ASMUSSEN Edited by P. GLYNN, T. MIKOSCH and T. ROLSKI Part 4. Simulation

More information

Recursive Least Squares for an Entropy Regularized MSE Cost Function

Recursive Least Squares for an Entropy Regularized MSE Cost Function Recursive Least Squares for an Entropy Regularized MSE Cost Function Deniz Erdogmus, Yadunandana N. Rao, Jose C. Principe Oscar Fontenla-Romero, Amparo Alonso-Betanzos Electrical Eng. Dept., University

More information

Two-sided bounds for L p -norms of combinations of products of independent random variables

Two-sided bounds for L p -norms of combinations of products of independent random variables Two-sided bounds for L p -norms of combinations of products of independent random ariables Ewa Damek (based on the joint work with Rafał Latała, Piotr Nayar and Tomasz Tkocz) Wrocław Uniersity, Uniersity

More information

Physics 111. Help sessions meet Sunday, 6:30-7:30 pm in CLIR Wednesday, 8-9 pm in NSC 098/099

Physics 111. Help sessions meet Sunday, 6:30-7:30 pm in CLIR Wednesday, 8-9 pm in NSC 098/099 ics Announcements day, ember 7, 2007 Ch 2: graphing - elocity s time graphs - acceleration s time graphs motion diagrams - acceleration Free Fall Kinematic Equations Structured Approach to Problem Soling

More information

The Kinetic Theory of Gases

The Kinetic Theory of Gases 978-1-107-1788-3 Classical and Quantum Thermal Physics The Kinetic Theory of Gases CHAPTER 1 1.0 Kinetic Theory, Classical and Quantum Thermodynamics Two important components of the unierse are: the matter

More information

Modeling Highway Traffic Volumes

Modeling Highway Traffic Volumes Modeling Highway Traffic Volumes Tomáš Šingliar1 and Miloš Hauskrecht 1 Computer Science Dept, Uniersity of Pittsburgh, Pittsburgh, PA 15260 {tomas, milos}@cs.pitt.edu Abstract. Most traffic management

More information

Prediction of anode arc root position in a DC arc plasma torch

Prediction of anode arc root position in a DC arc plasma torch Prediction of anode arc root position in a DC arc plasma torch He-Ping Li 1, E. Pfender 1, Xi Chen 1 Department of Mechanical Engineering, Uniersity of Minnesota, Minneapolis, MN 55455, USA Department

More information

Frequency Response Improvement in Microgrid Using Optimized VSG Control

Frequency Response Improvement in Microgrid Using Optimized VSG Control Frequency Response Improement in Microgrid Using Optimized SG Control B. Rathore, S. Chakrabarti, Senior Member, IEEE, S. Anand, Member, IEEE, Department of Electrical Engineering Indian Institute of Technology

More information

Decision-Point Signal to Noise Ratio (SNR)

Decision-Point Signal to Noise Ratio (SNR) Decision-Point Signal to Noise Ratio (SNR) Receiver Decision ^ SNR E E e y z Matched Filter Bound error signal at input to decision device Performance upper-bound on ISI channels Achieved on memoryless

More information

Least Squares with Examples in Signal Processing 1. 2 Overdetermined equations. 1 Notation. The sum of squares of x is denoted by x 2 2, i.e.

Least Squares with Examples in Signal Processing 1. 2 Overdetermined equations. 1 Notation. The sum of squares of x is denoted by x 2 2, i.e. Least Squares with Eamples in Signal Processing Ivan Selesnick March 7, 3 NYU-Poly These notes address (approimate) solutions to linear equations by least squares We deal with the easy case wherein the

More information

EE-597 Notes Quantization

EE-597 Notes Quantization EE-597 Notes Quantization Phil Schniter June, 4 Quantization Given a continuous-time and continuous-amplitude signal (t, processing and storage by modern digital hardware requires discretization in both

More information

To string together six theorems of physics by Pythagoras theorem

To string together six theorems of physics by Pythagoras theorem To string together six theorems of physics by Pythagoras theorem H. Y. Cui Department of Applied Physics Beijing Uniersity of Aeronautics and Astronautics Beijing, 00083, China ( May, 8, 2002 ) Abstract

More information

Single soliton solution to the extended KdV equation over uneven depth

Single soliton solution to the extended KdV equation over uneven depth Eur. Phys. J. E 7) : DOI./epje/i7-59-7 Regular Article THE EUROPEAN PHYSICAL JOURNAL E Single soliton solution to the etended KdV equation oer uneen depth George Rowlands, Piotr Rozmej,a, Eryk Infeld 3,

More information

SPECTRAL analysis of time-series recorded from the. Robust Estimation of Sparse Narrowband Spectra from Binary Neuronal Spiking Data

SPECTRAL analysis of time-series recorded from the. Robust Estimation of Sparse Narrowband Spectra from Binary Neuronal Spiking Data Robust Estimation of Sparse arrowband Spectra from Binary euronal Spiing Data Sina Miran, Patric L. Purdon, Emery. Brown, and Behtash Babadi Abstract Objectie: Characterizing the spectral properties of

More information

(x,y) 4. Calculus I: Differentiation

(x,y) 4. Calculus I: Differentiation 4. Calculus I: Differentiation 4. The eriatie of a function Suppose we are gien a cure with a point lying on it. If the cure is smooth at then we can fin a unique tangent to the cure at : If the tangent

More information

Performance Comparison of Two Implementations of the Leaky. LMS Adaptive Filter. Scott C. Douglas. University of Utah. Salt Lake City, Utah 84112

Performance Comparison of Two Implementations of the Leaky. LMS Adaptive Filter. Scott C. Douglas. University of Utah. Salt Lake City, Utah 84112 Performance Comparison of Two Implementations of the Leaky LMS Adaptive Filter Scott C. Douglas Department of Electrical Engineering University of Utah Salt Lake City, Utah 8411 Abstract{ The leaky LMS

More information

Variance Reduction for Stochastic Gradient Optimization

Variance Reduction for Stochastic Gradient Optimization Variance Reduction for Stochastic Gradient Optimization Chong Wang Xi Chen Alex Smola Eric P. Xing Carnegie Mellon Uniersity, Uniersity of California, Berkeley {chongw,xichen,epxing}@cs.cmu.edu alex@smola.org

More information

Towards Universal Cover Decoding

Towards Universal Cover Decoding International Symposium on Information Theory and its Applications, ISITA2008 Auckland, New Zealand, 7-10, December, 2008 Towards Uniersal Coer Decoding Nathan Axig, Deanna Dreher, Katherine Morrison,

More information

Motion in Two and Three Dimensions

Motion in Two and Three Dimensions PH 1-1D Spring 013 Motion in Two and Three Dimensions Lectures 5,6,7 Chapter 4 (Halliday/Resnick/Walker, Fundamentals of Physics 9 th edition) 1 Chapter 4 Motion in Two and Three Dimensions In this chapter

More information

v are uncorrelated, zero-mean, white

v are uncorrelated, zero-mean, white 6.0 EXENDED KALMAN FILER 6.1 Introduction One of the underlying assumptions of the Kalman filter is that it is designed to estimate the states of a linear system based on measurements that are a linear

More information

Analysis of universal adversarial perturbations

Analysis of universal adversarial perturbations 1 Analysis of uniersal adersarial perturbations Seyed-Mohsen Moosai-Dezfooli seyed.moosai@epfl.ch Alhussein Fawzi fawzi@cs.ucla.edu Omar Fawzi omar.fawzi@ens-lyon.fr ascal Frossard pascal.frossard@epfl.ch

More information

An Optimal Split-Plot Design for Performing a Mixture-Process Experiment

An Optimal Split-Plot Design for Performing a Mixture-Process Experiment Science Journal of Applied Mathematics and Statistics 217; 5(1): 15-23 http://www.sciencepublishinggroup.com/j/sjams doi: 1.11648/j.sjams.21751.13 ISSN: 2376-9491 (Print); ISSN: 2376-9513 (Online) An Optimal

More information

Macroscopic characterization with distribution functions. Physical Chemistry. Particles in a Box. Two equilibrium distributions

Macroscopic characterization with distribution functions. Physical Chemistry. Particles in a Box. Two equilibrium distributions Macroscopic characterization with distribution functions Physical Chemistry Lecture istributions and ransport Processes A macroscopic system contains a LARE number of particles, not all of which h the

More information

Information Propagation Speed in Bidirectional Vehicular Delay Tolerant Networks

Information Propagation Speed in Bidirectional Vehicular Delay Tolerant Networks Information Propagation Speed in Bidirectional Vehicular Delay Tolerant Networks Emmanuel Baccelli, Philippe Jacquet, Bernard Mans, Georgios Rodolakis INRIA, France, {firstname.lastname}@inria.fr Macquarie

More information

Unconstrained Multivariate Optimization

Unconstrained Multivariate Optimization Unconstrained Multivariate Optimization Multivariate optimization means optimization of a scalar function of a several variables: and has the general form: y = () min ( ) where () is a nonlinear scalar-valued

More information

A matrix Method for Interval Hermite Curve Segmentation O. Ismail, Senior Member, IEEE

A matrix Method for Interval Hermite Curve Segmentation O. Ismail, Senior Member, IEEE International Journal of Video&Image Processing Network Security IJVIPNS-IJENS Vol:15 No:03 7 A matrix Method for Interal Hermite Cure Segmentation O. Ismail, Senior Member, IEEE Abstract Since the use

More information