Grouped data clustering using a fast mixture-model-based algorithm
|
|
- Donald Willis
- 5 years ago
- Views:
Transcription
1 Poceedings of the 29 IEEE Intenational Confeence on Systems, Man, and Cybenetics San Antonio, TX, USA - Octobe 29 Gouped data clusteing using a fast mixtue-model-based algoithm Allou SAMÉ Laboatoie des Technologies Nouvelles Institut National de Recheche su les Tanspots et leu Sécuité Noisy-le-Gand, Fance same@inets.f Abstact Mixtue-model-based clusteing has become a popula appoach in many data analysis poblems fo its statistical popeties and the implementation simplicity of the EM algoithm. Howeve the computation time of the EM algoithm and its vaiants inceases significantly with the sample size. Fo lage data sets, pefoming clusteing on gouped data constitutes an efficient altenative to speed up the algoithms execution time. A apid and effective algoithm dedicated to gouped data clusteing is then poposed in this pape. Inspied by the Classification EM algoithm (), the poposed appoach estimates the missing sample at each iteation. An expeimental study using simulated data and eal acoustic emission data in the context of a flaw detection application on gas tans eveals good pefomances of the poposed appoach in tems of patitioning pecision and computing time. Index Tems Clusteing, mixtue models, EM algoithm, Classification EM algoithm, gouped Data, histogam I. INTRODUCTION Clusteing is an exploatoy data analysis technique which consists in identifying homogeneous goups in data and thus constitutes a cental poblematic in many applications including web data mining, bio-infomatics and image segmentation. Mixtue-model-based clusteing [1][6] continues to eceive inceasing attention fo its statistical popeties and its implementation simplicity. Its two commonly used appoaches [6] fo multidimensional data ae the maximum lielihood (ML) appoach and the classification maximum lielihood (CML) appoach. The ML appoach consists in estimating the model paamete by maximizing the log-lielihood by the Expectation Maximization (EM) algoithm and then to estimate a patition by the maximum a posteioi (MAP) ule. In the CML appoach, the model paamete and the patition ae simultaneously estimated by maximizing the classification loglielihood using the Classification Expectation Maximization () algoithm []. With the impovement of measuement devices and compute, data sets with a lage numbe of obsevations become fequent and the challenge is to develop clusteing methods whose execution time does not depend on the numbe of obsevations. Howeve the computation time of the EM and algoithms inceases significantly with the sample size. To speed up the EM algoithm, vaious attempts have been poposed: the incemental and spae veions poposed by Neal and Hinton [1], the scalable veion of the EM algoithm poposed by Badley et al. [2] to handle lage databases with a limited memoy buffe, the Mooe s appoach that summaizes data using the so-called multiesolution d-tee [9], the Ng and McLachlan appoach [11] that combines the incemental EM and the multiesolution d-tee appoach. Anothe geneic way of impoving the clusteing algoithms execution time is to opeate on gouped data o histogams instead of oiginal obsevations. This data stuctue also efeed as binned data [4] simply esults fom the conveion of oiginal obsevations into counts in disjoined subspaces of the oveall obsevations space. Gouped data can also natually occu when the measuement devices have finite pecision. Figue 1 shows example of gouped data esulting fom the discetization of eal acoustic emission data. McLachlan and Jones [8], and Cadez et al. [4] have theoetically fomulated the poblem of estimating a mixtue model fom gouped data as a missing data poblem, and have developed a dedicated EM algoithm. Howeve, the multivaiate fomulation of this algoithm equies the numeical evaluation of multiple integals which can be computationally expensive. Fig. 1. Gouped data esulting fom the discetization of eal acoustic emission data A faste mixtue-model-based algoithm fo gouped data clusteing is pesented in this pape. The poposed algoithm is inspied by the classification maximization lielihood appoach and incopoates a sample estimation at each iteation. The section 2 intoduces the essential concepts of gouped /9/$. c 29 IEEE /9/$. 29 IEEE 23
2 data in the famewo of mixtue models, and biefly ecalls the Cadez et al. appoach [4] fo paamete estimation via the EM algoithm. Ou clusteing appoach adapted to binned data is descibed in the thid section and an expeimental study using simulated data and eal acoustic emission data is summaized in Section 4. II. ESTIMATING A MIXTURE MODEL FROM GROUPED DATA The oiginal data ae epesented by an independent sample x (x 1,...,x n ) esulting fom a mixtue density of K components, defined on R p by K f(x i ; Φ) π f (x i ; θ ), (1) 1 whee Φ (π 1,...,π K, θ 1,...,θ K ), π 1,...,π K denotes the popotions of the mixtue and (θ 1,...,θ K ) the paamete of each component density. We denote by z (z 1,...,z n ) the set of classes associated to the sample x, wheez i {1,...,K}. In addition, we conside a patition (H 1,...,H v ) of R p into v hypeectangula bins and we assume that the only infomation we have petaining to the sample is the set of fequencies n of the obsevations x i belonging to the bins H, summaized by a vecto n (n 1,...,n v ) of counts, with v 1 n n. The obsevations of bin H can then be ewitten as (x ) 1 s n and thei associated classes labels by ( ) 1 s n. The estimation of the mixtue model paamete fom gouped data was fit intoduced by Dempste et al. [7] as a missing data poblem in the famewo of the EM algoithm, and then developed in the one-dimensional case by McLachlan and Jones [8]. It can be noticed that in this new situation, the missing data ae the oiginal obsevations and thei cluste membehip. The genealization to the multidimensional case can be attibuted to Cadez et al. [4]. To maximize the obseved log-lielihood defined by v L(Φ) logp(n; Φ) n log f(x; Φ)dx c, (2) 1 H c being a contant tem which does not depend on Φ, the EM algoithm stats fom an initial paamete Φ () and then altenates the two following steps until convegence : E-step (Expectation) Computation of the expectation of the complete data log-lielihood conditionally on the available data and the cuent paamete vecto Φ (q) : Q(Φ, Φ (q) )E[L c (Φ, x, z) n, Φ (q) ], whee the complete data log-lielihood is defined by L c (Φ, x, z) logp(n, x, z; Φ) K v n log π f (x ; θ ). 1 1 s1 with 1if, andothewise. M-step (Maximization) Computation of the paamete vecto Φ (q1) maximizing Q(Φ, Φ (q) ). Given the paamete vecto estimated pa the EM algoithm, a patition of the bins can easily be deived by assigning each bin H to the cluste which maximizes the posteio pobability that an obsevation x i of the bin H aises fom the th component of the mixtue. The main difficulty when implementing the EM algoithm applied to gouped data, is the numeical evaluation of multiples integals at each iteation, which can be computationally expensive. The following section intoduces a faste classification algoithm which does not equie integals calculations. III. THE PROPOSED CLASSIFICATION APPROACH The clusteing method intoduced hee, in the context of gouped data, is inspied by the Classification EM algoithm () [] applied to oiginal obsevations. The standad algoithm is an iteative clusteing algoithm yielding simultaneously the paamete and the classification. It iteatively maximizes, with espect to the component membehip vecto z (z 1,...,z n ) and the paamete vecto Φ, the classification log-lielihood citeion. The algoithm also can be intepeted as a classification veion of the EM algoithm []. In the famewo of gouped data, we popose an adaptation of the algoithm which consists in estimating simultaneously the missing data (x, z) and the paamete vecto Φ by maximizing the complete data log-lielihood which can be ewitten as v n L c (Φ, x, z) log π z f z (x ; θ z ). (3) 1 s1 The poposed algoithm stats with an initial paamete Φ () and then altenates, at the qth iteation, the two following steps until convegence: Step 1 Computation of (x (q1), z (q1) ) ag max (x,z) L c(φ (q), x, z) (4) Step 2 Computation of Φ (q1) agmax L c(φ, x (q1), z (q1) ) () Φ As in the classical veion of applied to oiginal obsevations, it can easily be poved that each iteation of this algoithms inceases the complete data log-lielihood citeion and that the convegence is obseved afte a finite numbe of iteations. The following sections detail the two steps of this algoithm unde the assumption of a Gaussian mixtue with diagonal covaiance matices which is a ealistic hypothesis fo many eal applications and moe paticulaly fo the application that will be intoduced in section 4. A. Step 1: maximization of the complete data log-lielihood with espect to the sample and the patition Unde the Gaussian mixtue assumption, the maximization of L c with espect to (x, z) is equivalent to the minimization 21
3 of h 1 (x, z) v n log Σ (q) 2logπ z (q) 1 s1 (x μ (q) ) T Σ 1 (x μ (q) ), (6) whee the π, μ and Σ ae espectively the popotions, mean vecto and covaiance matices of the mixtue. Minimizing h 1 is equivalent to minimize, fo all and s, h 2 (x, ) log Σ (q) 2logπ z (q) (x μ (q) ) T Σ 1 (x μ z (q) ) (7) unde the constaints x H and {1,...,K}. Since the minimization of h 2 depends only on the bin label, letus denote by (x (q1),z (q1) ) its solution. This solution can be obtained using the two following steps: (a) Fo 1...K, compute x (q1) (b) Set (x (q1) agmin(x μ (q) x H agmin,z (q1) )(x (q1) (x (q1) )T Σ (q) 1 (q) (x μ );, ),whee log Σ (q) 2logπ(q) μ (q) )T Σ (q) 1 (x (q1) μ (q1) ). The poblem (a) is a convex optimization poblem with linea constaints. It can be seen that x (q1) emains in the bounday of the bin H if μ (q1) is outside the bin H.Fo example, unde the assumption of a 2-dimensional Gaussian mixtue distibution with diagonal covaiance matices, if we set H 2 l1 [al ; bl [ and μ(q) (μ (q) 1,μ(q) 2 )T, it can be shown that x (q1) (x 1,x 2 )T with x 1 { a if μ (q) 1 <a μ (q) 1 b if a μ (q) 1 if μ (q) b x 2 1 >b { c if μ (q) 2 <c μ (q) 2 d if c μ (q) 2 d. if μ (q) 2 >d Figue 2 shows examples of locations of x in elation to the cuent value of μ. B. Step 2: maximization of the complete data log-lielihood with espect to the paamete vecto This step is simila to the M step of the standad algoithm [] applied to the sample (x (q1) 1,...,x v (q1) ) weighted by the fequencies (n 1,...,n v ). The popotions, mean vecto and covaiance matices which maximizes L c (Φ, x (q1), z (q1) ) ae given by π (q1) v 1 nz(q1) μ (q1) Σ (q1) diag n v 1 nz(q1) x (q1) v1 ( n z (q1) v 1 nz(q1) (x (q1) μ (q1) v 1 nz(q1) ) )(x (q1) μ (q1) ) T whee z 1if z, andothewise,anddiag(a) is the diagonal covaiance matix whose diagonal elements ae those of matix A. (a) (b) (c) Fig. 2. Exemples of localisations of x : fo cases (a) and (b) the Gaussian mean is outside the bin H and fo case (c) the Gaussian mean is inside the bin IV. EXPERIMENTAL STUDY This section evaluates the poposed algoithm, that we shall call Bin-, in tems of pecision and computing time using simulated data and eal wold data. A. Expeiments using simulated data The potocol of the simulations is as follows: n obsevations ae geneated accoding to a mixtue of 2 bivaiate Gaussian densities; these obsevations ae used to compute an histogam containing v bins (gouped data). Two diffeent mixtue models with % of theoetical Bayes eo ate have been consideed: mixtue model A: π 1 π 2 1/2, μ 1 ( 2, ), μ 2 (, ), Σ 1 Σ 2 diag(1, 1); mixtue model B: π 1 π 2 1/2, μ 1 (1.6, ), μ 2 (, ), Σ 1 diag(1, 1/8), Σ 2 diag(1/8, 1). Figue 3 shows, examples of simulations of gouped data accoding to mixtues A and B. Thee algoithm ae compaed in all these simulations: the standad algoithm applied to the oiginal data (individual obsevations), the Bin- and Bin-EM (EM applied to gouped data) algoithms applied to gouped data. The 22
4 bin Fig. 3. Example of a simulation of the gouped data accoding to model A (left) and B (ight) at bins bin quality of a patition povided by each algoithm is measued by computing its misclassification ate with espect to the simulated patition. Fo each simulated oiginal data set o gouped data set, the misclassification ates and CPU times ae aveaged ove diffeent samples (Monte-Calo simulation scheme). Figue 4 epots the misclassification ate obtained fo mixtues A and B in elation to the numbe of bins pe dimension, when n. We obseve a apid decease until the numbe of bins pe dimension eaches. T Theeafte, the misclassification ate is almost constant and close to that of : the thee algoithms become simila Fig.. CPU time in second fo the (no ma), bin-em (cicle) and bin- (squae) algoithms in elation to the sample size at bins fo mixtue B bin Model A bin Model B Fig. 4. Misclassification ate (%) in elation to the numbe of bins pe dimension, obtained fo n with (no ma), bin-em (cicle) and bin- (squae) fo mixtues A (top) and B (bottom) Figue epesents CPU times in second fo the thee algoithms with espect to the sample size fo a fixed numbe of bins pe dimension ( bins pe dimension). The execution time of Bin- and Bin-EM algoithms does not vay while that of the algoithm gows consideably with the sample size. Bin- outpefoms the and bin-em algoithms. B. Expeiments using eal acoustic emission data A motivation of this wo was to develop a compute-aided decision pocedue to assist the detection, in eal time, of damaged zones on the suface of a gas tan, though the use of acoustic emissions. Expets ae in ageement that spatial concentations of acoustic emissions, identified using spatial coodinates, ae of pimay impotance in the detection of damaged zones. Othe featues of acoustic emissions ae useful in distinguishing between majo and mino flaws. Ou method, theefoe, consists of two steps : Identification of zones whee acoustic emissions ae concentated (clusteing step) which is the object of the pesent study; Sepaation of the identified cluste into diffeent categoies accoding to the seveity of the impefection: these categoies ae temed mino, active and citical. Thisstep is done using standad discimination methods such as Bayesian discimination with Gaussian densities. Accoding to specialists in the field, the method we have descibed poduces satisfactoy esults using the algoithm fo the fit step, so long as the numbe of acoustic emissions does not exceed 1. When thee ae moe than 1 emissions the clusteing step becomes too slow (moe than a few seconds delay) fo a eal-time application. The assumption of Gaussian mixtue with diagonal covaiance matices can be attibuted to the fact that the damage zones ae usually to be found hoizontally and vetically along welding lines. To evaluate ou new stategy, we pefomed a compaison of and Bin- on a eal data sets of acoustic emissions. The numbe of cluste is selected by maximizing the integated classification lielihood citeion (ICL) [3] 23
5 which is the complete lielihood penalized by the numbe of fee paamete of the mixtue model. Both stategies, using and bin-, select a model with 11 cluste. Table I shows the misclassification ate between and Bin- in elation to the numbe of bins. Figue 6 shows the esulting patitions given by and Bin- at bins pe dimension and 8 bins pe dimension. TABLE I MISCLASSIFICATION RATES BETWEEN AND BIN- FOR REAL ACOUSTIC EMISSION DATA Numbe of bins Vs. Bin-.7% 6 6.8% % % % This shows that esults obtained with Bin- ae close to those of the algoithm applied to the oiginal sample. In this application, we have theefoe achieved ou aim of obtaining patitions simila to patitions but in less time. V. CONCLUSION A new clusteing algoithm fo gouped data has been poposed in this pape. The main objective was to obtain a apid and effective algoithm fo clusteing low-dimensional massive data sets. The poposed Bin- clusteing appoach poduces patitions fom gouped data, which ae close to those computed by the algoithm applied to the oiginal obsevations. Almost no diffeences in esults ae obseved between and Bin- patitions when the numbe of bins is sufficiently lage (geate than in the simulations we have pesented). The execution time of the poposed algoithm is almost constant, since it depends only on the numbe of bins, while the execution time fo inceases significantly as the numbe of available obsevations inceases. The application of the poposed algoithm on eal acoustic emission data has also eveal good pefomances of the poposed algoithm which theefoe epesents an efficient altenative fo clusteing lage data sets. REFERENCES [1] J.D. Banfield and A.E. Raftey, Model-based Gaussian and non Gaussian clusteing. Biometics, 49, , 23. [2] P.S. Badley, M.U. Fayyad and C.A. Reina, Scaling EM (Expectation Maximization) clusteing to lage databases. Technical Reppot MSR- TR-98-, Micosoft Reseach, [3] C. Bienaci and G. Celeux and G. Govaet, Assessing a mixtue model fo clusteing with the integated completed lielihood. IEEE Tansactions on Patten Analysis and Machine Intelligence. 22, 7, 719-7, 2. [4] I.V. Cadez and P. Smyth and G.J. McLachlan and C.E. McLaen, Maximum lielihood estimation of mixtue densities fo binned and tuncated multivaiate data. Machine Leaning, 47, 7-34, 2. [] G. Celeux and G. Govaet, A classification EM algoithm fo clusteing and two stochastic veions. Computational Statistics and Data Analysis, 14, 3-332, [6] G. Celeux and G. Govaet, Gaussian paimonious clusteing models. Patten Recognition, 28(), , 199. [7] A.P. Dempste and N.M. Laid and D.B. Rubin, Maximum lielihood fom incomplete data via the EM algoithm. Jounal of the Royal Statistical Society, Seies B, 39(1), 1-38, (oiginal sample) bins bins Fig. 6. Results obtained with (top) and Bin- (middle and bottom) on eal acoustic emission data [8] G.J. McLachlan and P.N. Jones, Fitting mixtue models to gouped and tuncated data via the EM algoithm. Biometics, 44(2), 71-78, [9] A. W. Mooe, Vey fast EM based mixtue model clusteing using multiesolution d-tees. In Solla M.S. Keans and D.A Cohn, edito, Advances in Neual Infomation Pocessing Systems, 11, 43-49, MIT Pess, [1] R.M. Neal and Hinton, A view of the EM algoithm that justifies incemental, spae and othe vaiants. In M.I. Jodan, edito, Leaning in Gaphical Models, -368, Kluwe, Dodecht, 1998 [11] S.-K. Ng and G. J. McLachlan, On some vaiants of the EM algoithm fo the finite mixtue models. Autian Jounal of Statistics, 32(1-2), ,
Pearson s Chi-Square Test Modifications for Comparison of Unweighted and Weighted Histograms and Two Weighted Histograms
Peason s Chi-Squae Test Modifications fo Compaison of Unweighted and Weighted Histogams and Two Weighted Histogams Univesity of Akueyi, Bogi, v/noduslód, IS-6 Akueyi, Iceland E-mail: nikolai@unak.is Two
More informationCentral Coverage Bayes Prediction Intervals for the Generalized Pareto Distribution
Statistics Reseach Lettes Vol. Iss., Novembe Cental Coveage Bayes Pediction Intevals fo the Genealized Paeto Distibution Gyan Pakash Depatment of Community Medicine S. N. Medical College, Aga, U. P., India
More informationAPPLICATION OF MAC IN THE FREQUENCY DOMAIN
PPLICION OF MC IN HE FREQUENCY DOMIN D. Fotsch and D. J. Ewins Dynamics Section, Mechanical Engineeing Depatment Impeial College of Science, echnology and Medicine London SW7 2B, United Kingdom BSRC he
More informationTemporal-Difference Learning
.997 Decision-Making in Lage-Scale Systems Mach 17 MIT, Sping 004 Handout #17 Lectue Note 13 1 Tempoal-Diffeence Leaning We now conside the poblem of computing an appopiate paamete, so that, given an appoximation
More informationMULTILAYER PERCEPTRONS
Last updated: Nov 26, 2012 MULTILAYER PERCEPTRONS Outline 2 Combining Linea Classifies Leaning Paametes Outline 3 Combining Linea Classifies Leaning Paametes Implementing Logical Relations 4 AND and OR
More informationCOMPUTATIONS OF ELECTROMAGNETIC FIELDS RADIATED FROM COMPLEX LIGHTNING CHANNELS
Pogess In Electomagnetics Reseach, PIER 73, 93 105, 2007 COMPUTATIONS OF ELECTROMAGNETIC FIELDS RADIATED FROM COMPLEX LIGHTNING CHANNELS T.-X. Song, Y.-H. Liu, and J.-M. Xiong School of Mechanical Engineeing
More informationGraphs of Sine and Cosine Functions
Gaphs of Sine and Cosine Functions In pevious sections, we defined the tigonometic o cicula functions in tems of the movement of a point aound the cicumfeence of a unit cicle, o the angle fomed by the
More informationPsychometric Methods: Theory into Practice Larry R. Price
ERRATA Psychometic Methods: Theoy into Pactice Lay R. Pice Eos wee made in Equations 3.5a and 3.5b, Figue 3., equations and text on pages 76 80, and Table 9.1. Vesions of the elevant pages that include
More informationDuality between Statical and Kinematical Engineering Systems
Pape 00, Civil-Comp Ltd., Stiling, Scotland Poceedings of the Sixth Intenational Confeence on Computational Stuctues Technology, B.H.V. Topping and Z. Bittna (Editos), Civil-Comp Pess, Stiling, Scotland.
More informationBayesian Analysis of Topp-Leone Distribution under Different Loss Functions and Different Priors
J. tat. Appl. Po. Lett. 3, No. 3, 9-8 (6) 9 http://dx.doi.og/.8576/jsapl/33 Bayesian Analysis of Topp-Leone Distibution unde Diffeent Loss Functions and Diffeent Pios Hummaa ultan * and. P. Ahmad Depatment
More informationInteraction of Feedforward and Feedback Streams in Visual Cortex in a Firing-Rate Model of Columnar Computations. ( r)
Supplementay mateial fo Inteaction of Feedfowad and Feedback Steams in Visual Cotex in a Fiing-Rate Model of Columna Computations Tobias Bosch and Heiko Neumann Institute fo Neual Infomation Pocessing
More information4/18/2005. Statistical Learning Theory
Statistical Leaning Theoy Statistical Leaning Theoy A model of supevised leaning consists of: a Envionment - Supplying a vecto x with a fixed but unknown pdf F x (x b Teache. It povides a desied esponse
More informationAn extended target tracking method with random finite set observations
4th Intenational Confeence on Infomation Fusion Chicago Illinois USA July 5-8 0 An extended taget tacing method with andom finite set obsevations Hongyan Zhu Chongzhao Han Chen Li Dept. of Electonic &
More informationMultiple Experts with Binary Features
Multiple Expets with Binay Featues Ye Jin & Lingen Zhang Decembe 9, 2010 1 Intoduction Ou intuition fo the poect comes fom the pape Supevised Leaning fom Multiple Expets: Whom to tust when eveyone lies
More informationPulse Neutron Neutron (PNN) tool logging for porosity Some theoretical aspects
Pulse Neuton Neuton (PNN) tool logging fo poosity Some theoetical aspects Intoduction Pehaps the most citicism of Pulse Neuton Neuon (PNN) logging methods has been chage that PNN is to sensitive to the
More informationState tracking control for Takagi-Sugeno models
State tacing contol fo Taagi-Sugeno models Souad Bezzaoucha, Benoît Max,3,DidieMaquin,3 and José Ragot,3 Abstact This wo addesses the model efeence tacing contol poblem It aims to highlight the encouteed
More informationAlternative Tests for the Poisson Distribution
Chiang Mai J Sci 015; 4() : 774-78 http://epgsciencecmuacth/ejounal/ Contibuted Pape Altenative Tests fo the Poisson Distibution Manad Khamkong*[a] and Pachitjianut Siipanich [b] [a] Depatment of Statistics,
More informationHydroelastic Analysis of a 1900 TEU Container Ship Using Finite Element and Boundary Element Methods
TEAM 2007, Sept. 10-13, 2007,Yokohama, Japan Hydoelastic Analysis of a 1900 TEU Containe Ship Using Finite Element and Bounday Element Methods Ahmet Egin 1)*, Levent Kaydıhan 2) and Bahadı Uğulu 3) 1)
More informationLight Time Delay and Apparent Position
Light Time Delay and ppaent Position nalytical Gaphics, Inc. www.agi.com info@agi.com 610.981.8000 800.220.4785 Contents Intoduction... 3 Computing Light Time Delay... 3 Tansmission fom to... 4 Reception
More informationarxiv: v2 [physics.data-an] 15 Jul 2015
Limitation of the Least Squae Method in the Evaluation of Dimension of Factal Bownian Motions BINGQIANG QIAO,, SIMING LIU, OUDUN ZENG, XIANG LI, and BENZONG DAI Depatment of Physics, Yunnan Univesity,
More informationA New Method of Estimation of Size-Biased Generalized Logarithmic Series Distribution
The Open Statistics and Pobability Jounal, 9,, - A New Method of Estimation of Size-Bied Genealized Logaithmic Seies Distibution Open Access Khushid Ahmad Mi * Depatment of Statistics, Govt Degee College
More informationC e f paamete adaptation f (' x) ' ' d _ d ; ; e _e K p K v u ^M() RBF NN ^h( ) _ obot s _ s n W ' f x x xm xm f x xm d Figue : Block diagam of comput
A Neual-Netwok Compensato with Fuzzy Robustication Tems fo Impoved Design of Adaptive Contol of Robot Manipulatos Y.H. FUNG and S.K. TSO Cente fo Intelligent Design, Automation and Manufactuing City Univesity
More informationChapter 3 Optical Systems with Annular Pupils
Chapte 3 Optical Systems with Annula Pupils 3 INTRODUCTION In this chapte, we discuss the imaging popeties of a system with an annula pupil in a manne simila to those fo a system with a cicula pupil The
More informationInformation Retrieval Advanced IR models. Luca Bondi
Advanced IR models Luca Bondi Advanced IR models 2 (LSI) Pobabilistic Latent Semantic Analysis (plsa) Vecto Space Model 3 Stating point: Vecto Space Model Documents and queies epesented as vectos in the
More informationMONTE CARLO STUDY OF PARTICLE TRANSPORT PROBLEM IN AIR POLLUTION. R. J. Papancheva, T. V. Gurov, I. T. Dimov
Pliska Stud. Math. Bulga. 14 (23), 17 116 STUDIA MATHEMATICA BULGARICA MOTE CARLO STUDY OF PARTICLE TRASPORT PROBLEM I AIR POLLUTIO R. J. Papancheva, T. V. Guov, I. T. Dimov Abstact. The actual tanspot
More informationEncapsulation theory: the transformation equations of absolute information hiding.
1 Encapsulation theoy: the tansfomation equations of absolute infomation hiding. Edmund Kiwan * www.edmundkiwan.com Abstact This pape descibes how the potential coupling of a set vaies as the set is tansfomed,
More information3.1 Random variables
3 Chapte III Random Vaiables 3 Random vaiables A sample space S may be difficult to descibe if the elements of S ae not numbes discuss how we can use a ule by which an element s of S may be associated
More informationChapter 5 Linear Equations: Basic Theory and Practice
Chapte 5 inea Equations: Basic Theoy and actice In this chapte and the next, we ae inteested in the linea algebaic equation AX = b, (5-1) whee A is an m n matix, X is an n 1 vecto to be solved fo, and
More informationOn Utilization of K-Means for Determination
Jounal of Softwae Engineeing and Applications, 07, 0, 605-64 http://www.scip.og/jounal/jsea ISSN Online: 945-34 ISSN Pint: 945-36 On Utilization of K-Means fo Detemination of q-paamete fo Tsallis-Entopy-Maximized-FCM
More informationMitscherlich s Law: Sum of two exponential Processes; Conclusions 2009, 1 st July
Mitschelich s Law: Sum of two exponential Pocesses; Conclusions 29, st July Hans Schneebege Institute of Statistics, Univesity of Elangen-Nünbeg, Gemany Summay It will be shown, that Mitschelich s fomula,
More informationAbsorption Rate into a Small Sphere for a Diffusing Particle Confined in a Large Sphere
Applied Mathematics, 06, 7, 709-70 Published Online Apil 06 in SciRes. http://www.scip.og/jounal/am http://dx.doi.og/0.46/am.06.77065 Absoption Rate into a Small Sphee fo a Diffusing Paticle Confined in
More informationA NEW VARIABLE STIFFNESS SPRING USING A PRESTRESSED MECHANISM
Poceedings of the ASME 2010 Intenational Design Engineeing Technical Confeences & Computes and Infomation in Engineeing Confeence IDETC/CIE 2010 August 15-18, 2010, Monteal, Quebec, Canada DETC2010-28496
More informationA matrix method based on the Fibonacci polynomials to the generalized pantograph equations with functional arguments
A mati method based on the Fibonacci polynomials to the genealized pantogaph equations with functional aguments Ayşe Betül Koç*,a, Musa Çama b, Aydın Kunaz a * Coespondence: aysebetuloc @ selcu.edu.t a
More informationHammerstein Model Identification Based On Instrumental Variable and Least Square Methods
Intenational Jounal of Emeging Tends & Technology in Compute Science (IJETTCS) Volume 2, Issue, Januay Febuay 23 ISSN 2278-6856 Hammestein Model Identification Based On Instumental Vaiable and Least Squae
More informationSurveillance Points in High Dimensional Spaces
Société de Calcul Mathématique SA Tools fo decision help since 995 Suveillance Points in High Dimensional Spaces by Benad Beauzamy Januay 06 Abstact Let us conside any compute softwae, elying upon a lage
More informationGoodness-of-fit for composite hypotheses.
Section 11 Goodness-of-fit fo composite hypotheses. Example. Let us conside a Matlab example. Let us geneate 50 obsevations fom N(1, 2): X=nomnd(1,2,50,1); Then, unning a chi-squaed goodness-of-fit test
More informationApplication of homotopy perturbation method to the Navier-Stokes equations in cylindrical coordinates
Computational Ecology and Softwae 5 5(): 9-5 Aticle Application of homotopy petubation method to the Navie-Stokes equations in cylindical coodinates H. A. Wahab Anwa Jamal Saia Bhatti Muhammad Naeem Muhammad
More informationThe Substring Search Problem
The Substing Seach Poblem One algoithm which is used in a vaiety of applications is the family of substing seach algoithms. These algoithms allow a use to detemine if, given two chaacte stings, one is
More informationWeb-based Supplementary Materials for. Controlling False Discoveries in Multidimensional Directional Decisions, with
Web-based Supplementay Mateials fo Contolling False Discoveies in Multidimensional Diectional Decisions, with Applications to Gene Expession Data on Odeed Categoies Wenge Guo Biostatistics Banch, National
More informationJ. Electrical Systems 1-3 (2005): Regular paper
K. Saii D. Rahem S. Saii A Miaoui Regula pape Coupled Analytical-Finite Element Methods fo Linea Electomagnetic Actuato Analysis JES Jounal of Electical Systems In this pape, a linea electomagnetic actuato
More informationModeling and Calculation of Optical Amplification in One Dimensional Case of Laser Medium Using Finite Difference Time Domain Method
Jounal of Physics: Confeence Seies PAPER OPEN ACCESS Modeling and Calculation of Optical Amplification in One Dimensional Case of Lase Medium Using Finite Diffeence Time Domain Method To cite this aticle:
More informationDetermining solar characteristics using planetary data
Detemining sola chaacteistics using planetay data Intoduction The Sun is a G-type main sequence sta at the cente of the Sola System aound which the planets, including ou Eath, obit. In this investigation
More informationAnalytical Solutions for Confined Aquifers with non constant Pumping using Computer Algebra
Poceedings of the 006 IASME/SEAS Int. Conf. on ate Resouces, Hydaulics & Hydology, Chalkida, Geece, May -3, 006 (pp7-) Analytical Solutions fo Confined Aquifes with non constant Pumping using Compute Algeba
More informationGradient-based Neural Network for Online Solution of Lyapunov Matrix Equation with Li Activation Function
Intenational Confeence on Infomation echnology and Management Innovation (ICIMI 05) Gadient-based Neual Netwok fo Online Solution of Lyapunov Matix Equation with Li Activation unction Shiheng Wang, Shidong
More informationA scaling-up methodology for co-rotating twin-screw extruders
A scaling-up methodology fo co-otating twin-scew extudes A. Gaspa-Cunha, J. A. Covas Institute fo Polymes and Composites/I3N, Univesity of Minho, Guimaães 4800-058, Potugal Abstact. Scaling-up of co-otating
More informationINFLUENCE OF GROUND INHOMOGENEITY ON WIND INDUCED GROUND VIBRATIONS. Abstract
INFLUENCE OF GROUND INHOMOGENEITY ON WIND INDUCED GROUND VIBRATIONS Mohammad Mohammadi, National Cente fo Physical Acoustics, Univesity of Mississippi, MS Caig J. Hicey, National Cente fo Physical Acoustics,
More informationRigid Body Dynamics 2. CSE169: Computer Animation Instructor: Steve Rotenberg UCSD, Winter 2018
Rigid Body Dynamics 2 CSE169: Compute Animation nstucto: Steve Rotenbeg UCSD, Winte 2018 Coss Poduct & Hat Opeato Deivative of a Rotating Vecto Let s say that vecto is otating aound the oigin, maintaining
More informationGaussian proposal density using moment matching in SMC methods
Stat Comput (009 19: 03 08 DOI.07/s11-008-9084-9 Gaussian poposal density using moment matching in SMC methods S. Saha P.K. Mandal Y. Boes H. Diessen A. Bagchi Received: 3 Apil 007 / Accepted: 30 June
More informationEFFECTS OF FRINGING FIELDS ON SINGLE PARTICLE DYNAMICS. M. Bassetti and C. Biscari INFN-LNF, CP 13, Frascati (RM), Italy
Fascati Physics Seies Vol. X (998), pp. 47-54 4 th Advanced ICFA Beam Dynamics Wokshop, Fascati, Oct. -5, 997 EFFECTS OF FRININ FIELDS ON SINLE PARTICLE DYNAMICS M. Bassetti and C. Biscai INFN-LNF, CP
More informationFresnel Diffraction. monchromatic light source
Fesnel Diffaction Equipment Helium-Neon lase (632.8 nm) on 2 axis tanslation stage, Concave lens (focal length 3.80 cm) mounted on slide holde, iis mounted on slide holde, m optical bench, micoscope slide
More informationReconstruction of 3D Anisotropic Objects by VIE and Model-Based Inversion Methods
Pogess In Electomagnetics Reseach C, Vol. 8, 97, 08 Reconstuction of D Anisotopic Objects by VIE and Model-Based Invesion Methods Lin E. Sun, * and Mei Song Tong Abstact A model-based invesion algoithm
More informationMathematical Model of Magnetometric Resistivity. Sounding for a Conductive Host. with a Bulge Overburden
Applied Mathematical Sciences, Vol. 7, 13, no. 7, 335-348 Mathematical Model of Magnetometic Resistivity Sounding fo a Conductive Host with a Bulge Ovebuden Teeasak Chaladgan Depatment of Mathematics Faculty
More informationExperiment I Voltage Variation and Control
ELE303 Electicity Netwoks Expeiment I oltage aiation and ontol Objective To demonstate that the voltage diffeence between the sending end of a tansmission line and the load o eceiving end depends mainly
More informationPredicting Cone-in-Cone Blender Efficiencies from Key Material Properties
Pedicting Cone-in-Cone Blende Efficiencies fom Key Mateial Popeties By: D. Key Johanson Mateial Flow Solutions, Inc. NOTICE: This is the autho s vesion of a wok accepted fo publication by Elsevie. Changes
More informationA Multivariate Normal Law for Turing s Formulae
A Multivaiate Nomal Law fo Tuing s Fomulae Zhiyi Zhang Depatment of Mathematics and Statistics Univesity of Noth Caolina at Chalotte Chalotte, NC 28223 Abstact This pape establishes a sufficient condition
More informationBounds on the performance of back-to-front airplane boarding policies
Bounds on the pefomance of bac-to-font aiplane boading policies Eitan Bachmat Michael Elin Abstact We povide bounds on the pefomance of bac-to-font aiplane boading policies. In paticula, we show that no
More informationSwissmetro: design methods for ironless linear transformer
Swissmeto: design methods fo ionless linea tansfome Nicolas Macabey GESTE Engineeing SA Scientific Pak PSE-C, CH-05 Lausanne, Switzeland Tel (+4) 2 693 83 60, Fax. (+4) 2 693 83 6, nicolas.macabey@geste.ch
More informationTo Feel a Force Chapter 7 Static equilibrium - torque and friction
To eel a oce Chapte 7 Chapte 7: Static fiction, toque and static equilibium A. Review of foce vectos Between the eath and a small mass, gavitational foces of equal magnitude and opposite diection act on
More informationIdentification of the degradation of railway ballast under a concrete sleeper
Identification of the degadation of ailway ballast unde a concete sleepe Qin Hu 1) and Heung Fai Lam ) 1), ) Depatment of Civil and Achitectual Engineeing, City Univesity of Hong Kong, Hong Kong SAR, China.
More informationStanford University CS259Q: Quantum Computing Handout 8 Luca Trevisan October 18, 2012
Stanfod Univesity CS59Q: Quantum Computing Handout 8 Luca Tevisan Octobe 8, 0 Lectue 8 In which we use the quantum Fouie tansfom to solve the peiod-finding poblem. The Peiod Finding Poblem Let f : {0,...,
More informationA Comment on Increasing Returns and Spatial. Unemployment Disparities
The Society fo conomic Studies The nivesity of Kitakyushu Woking Pape Seies No.06-5 (accepted in Mach, 07) A Comment on Inceasing Retuns and Spatial nemployment Dispaities Jumpei Tanaka ** The nivesity
More informationControl Chart Analysis of E k /M/1 Queueing Model
Intenational OPEN ACCESS Jounal Of Moden Engineeing Reseach (IJMER Contol Chat Analysis of E /M/1 Queueing Model T.Poongodi 1, D. (Ms. S. Muthulashmi 1, (Assistant Pofesso, Faculty of Engineeing, Pofesso,
More informationFUSE Fusion Utility Sequence Estimator
FUSE Fusion Utility Sequence Estimato Belu V. Dasaathy Dynetics, Inc. P. O. Box 5500 Huntsville, AL 3584-5500 belu.d@dynetics.com Sean D. Townsend Dynetics, Inc. P. O. Box 5500 Huntsville, AL 3584-5500
More informationEstimation of the Correlation Coefficient for a Bivariate Normal Distribution with Missing Data
Kasetsat J. (Nat. Sci. 45 : 736-74 ( Estimation of the Coelation Coefficient fo a Bivaiate Nomal Distibution with Missing Data Juthaphon Sinsomboonthong* ABSTRACT This study poposes an estimato of the
More informationA Comparative Study of Exponential Time between Events Charts
Quality Technology & Quantitative Management Vol. 3, No. 3, pp. 347-359, 26 QTQM ICAQM 26 A Compaative Study of Exponential Time between Events Chats J. Y. Liu 1, M. Xie 1, T. N. Goh 1 and P. R. Shama
More informationQuasi-Randomness and the Distribution of Copies of a Fixed Graph
Quasi-Randomness and the Distibution of Copies of a Fixed Gaph Asaf Shapia Abstact We show that if a gaph G has the popety that all subsets of vetices of size n/4 contain the coect numbe of tiangles one
More informationHOW TO TEACH THE FUNDAMENTALS OF INFORMATION SCIENCE, CODING, DECODING AND NUMBER SYSTEMS?
6th INTERNATIONAL MULTIDISCIPLINARY CONFERENCE HOW TO TEACH THE FUNDAMENTALS OF INFORMATION SCIENCE, CODING, DECODING AND NUMBER SYSTEMS? Cecília Sitkuné Göömbei College of Nyíegyháza Hungay Abstact: The
More informationASTR415: Problem Set #6
ASTR45: Poblem Set #6 Cuan D. Muhlbege Univesity of Mayland (Dated: May 7, 27) Using existing implementations of the leapfog and Runge-Kutta methods fo solving coupled odinay diffeential equations, seveal
More informationMultiple Criteria Secretary Problem: A New Approach
J. Stat. Appl. Po. 3, o., 9-38 (04 9 Jounal of Statistics Applications & Pobability An Intenational Jounal http://dx.doi.og/0.785/jsap/0303 Multiple Citeia Secetay Poblem: A ew Appoach Alaka Padhye, and
More information7.2. Coulomb s Law. The Electric Force
Coulomb s aw Recall that chaged objects attact some objects and epel othes at a distance, without making any contact with those objects Electic foce,, o the foce acting between two chaged objects, is somewhat
More informationGENERALIZED STATISTICAL METHODS FOR UNSUPERVISED MINORITY CLASS DETECTION IN MIXED DATA SETS. Uwe F. Mayer
GENERALIZED STATISTICAL METHODS FOR UNSUPERVISED MINORITY CLASS DETECTION IN MIXED DATA SETS Cécile Levasseu Jacobs School of Engineeing Univesity of Califonia, San Diego La Jolla, CA, USA clevasseu@ucsd.edu
More informationAdaptive Checkpointing in Dynamic Grids for Uncertain Job Durations
Adaptive Checkpointing in Dynamic Gids fo Uncetain Job Duations Maia Chtepen, Bat Dhoedt, Filip De Tuck, Piet Demeeste NTEC-BBT, Ghent Univesity, Sint-Pietesnieuwstaat 41, Ghent, Belgium {maia.chtepen,
More informationLET a random variable x follows the two - parameter
INTERNATIONAL JOURNAL OF MATHEMATICS AND SCIENTIFIC COMPUTING ISSN: 2231-5330, VOL. 5, NO. 1, 2015 19 Shinkage Bayesian Appoach in Item - Failue Gamma Data In Pesence of Pio Point Guess Value Gyan Pakash
More informationLecture 10. Vertical coordinates General vertical coordinate
Lectue 10 Vetical coodinates We have exclusively used height as the vetical coodinate but thee ae altenative vetical coodinates in use in ocean models, most notably the teainfollowing coodinate models
More information6 PROBABILITY GENERATING FUNCTIONS
6 PROBABILITY GENERATING FUNCTIONS Cetain deivations pesented in this couse have been somewhat heavy on algeba. Fo example, detemining the expectation of the Binomial distibution (page 5.1 tuned out to
More informationLinear Program for Partially Observable Markov Decision Processes. MS&E 339B June 9th, 2004 Erick Delage
Linea Pogam fo Patiall Obsevable Makov Decision Pocesses MS&E 339B June 9th 2004 Eick Delage Intoduction Patiall Obsevable Makov Decision Pocesses Etension of the Makov Decision Pocess to a wold with uncetaint
More informationNew problems in universal algebraic geometry illustrated by boolean equations
New poblems in univesal algebaic geomety illustated by boolean equations axiv:1611.00152v2 [math.ra] 25 Nov 2016 Atem N. Shevlyakov Novembe 28, 2016 Abstact We discuss new poblems in univesal algebaic
More informationMONTE CARLO SIMULATION OF FLUID FLOW
MONTE CARLO SIMULATION OF FLUID FLOW M. Ragheb 3/7/3 INTRODUCTION We conside the situation of Fee Molecula Collisionless and Reflective Flow. Collisionless flows occu in the field of aefied gas dynamics.
More informationCALCULATING THE NUMBER OF TWIN PRIMES WITH SPECIFIED DISTANCE BETWEEN THEM BASED ON THE SIMPLEST PROBABILISTIC MODEL
U.P.B. Sci. Bull. Seies A, Vol. 80, Iss.3, 018 ISSN 13-707 CALCULATING THE NUMBER OF TWIN PRIMES WITH SPECIFIED DISTANCE BETWEEN THEM BASED ON THE SIMPLEST PROBABILISTIC MODEL Sasengali ABDYMANAPOV 1,
More informationarxiv: v2 [astro-ph] 16 May 2008
New Anomalies in Cosmic Micowave Backgound Anisotopy: Violation of the Isotopic Gaussian Hypothesis in Low-l Modes Shi Chun, Su and M.-C., Chu Depatment of Physics and Institute of Theoetical Physics,
More informationLiquid gas interface under hydrostatic pressure
Advances in Fluid Mechanics IX 5 Liquid gas inteface unde hydostatic pessue A. Gajewski Bialystok Univesity of Technology, Faculty of Civil Engineeing and Envionmental Engineeing, Depatment of Heat Engineeing,
More information1 Similarity Analysis
ME43A/538A/538B Axisymmetic Tubulent Jet 9 Novembe 28 Similaity Analysis. Intoduction Conside the sketch of an axisymmetic, tubulent jet in Figue. Assume that measuements of the downsteam aveage axial
More informationModeling of High Temperature Superconducting Tapes, Arrays and AC Cables Using COMSOL
Except fom the Poceedings of the COMSOL Confeence 2010 Pais Modeling of High Tempeatue Supeconducting Tapes, Aays and AC Cables Using COMSOL Oleg Chevtchenko * Technical Univesity of Delft, The Nethelands
More informationGENLOG Multinomial Loglinear and Logit Models
GENLOG Multinomial Loglinea and Logit Models Notation This chapte descibes the algoithms used to calculate maximum-likelihood estimates fo the multinomial loglinea model and the multinomial logit model.
More informationCSCE 478/878 Lecture 4: Experimental Design and Analysis. Stephen Scott. 3 Building a tree on the training set Introduction. Outline.
In Homewok, you ae (supposedly) Choosing a data set 2 Extacting a test set of size > 3 3 Building a tee on the taining set 4 Testing on the test set 5 Repoting the accuacy (Adapted fom Ethem Alpaydin and
More informationComputational Methods of Solid Mechanics. Project report
Computational Methods of Solid Mechanics Poject epot Due on Dec. 6, 25 Pof. Allan F. Bowe Weilin Deng Simulation of adhesive contact with molecula potential Poject desciption In the poject, we will investigate
More informationThe Persistence of Most Probable Explanations in Bayesian Networks
EAI 2014 T. Schaub et al. (Eds.) 2014 The Authos and IOS Pess. This aticle is published online with Open Access by IOS Pess and distibuted unde the tems of the eative ommons Attibution Non-ommecial License.
More informationAST 121S: The origin and evolution of the Universe. Introduction to Mathematical Handout 1
Please ead this fist... AST S: The oigin and evolution of the Univese Intoduction to Mathematical Handout This is an unusually long hand-out and one which uses in places mathematics that you may not be
More information1D2G - Numerical solution of the neutron diffusion equation
DG - Numeical solution of the neuton diffusion equation Y. Danon Daft: /6/09 Oveview A simple numeical solution of the neuton diffusion equation in one dimension and two enegy goups was implemented. Both
More informationIMA Preprint Series # 2174
TRANSLATED POISSON MIXTURE MODEL FOR STRATIFICATION LEARNING By Gloia Hao Gegoy Randal and Guillemo Sapio IMA Pepint Seies # 2174 ( Septembe 27 ) INSTITUTE FOR MATHEMATICS AND ITS APPLICATIONS UNIVERSITY
More informationBasic Bridge Circuits
AN7 Datafoth Copoation Page of 6 DID YOU KNOW? Samuel Hunte Chistie (784-865) was bon in London the son of James Chistie, who founded Chistie's Fine At Auctionees. Samuel studied mathematics at Tinity
More informationMulti-Objective Optimization Algorithms for Finite Element Model Updating
Multi-Objective Optimization Algoithms fo Finite Element Model Updating E. Ntotsios, C. Papadimitiou Univesity of Thessaly Geece Outline STRUCTURAL IDENTIFICATION USING MEASURED MODAL DATA Weighted Modal
More informationMaking Fisher Discriminant Analysis Scalable
Making Fishe Disciminant Analysis Scalable Bojun Tu TUBOJUN@GMAIL.COM Depatment of Compute Science & Engineeing, Shanghai Jiao Tong Univesity, Shanghai, China Zhihua Zhang ZHIHUA@SJTU.EDU.CN Depatment
More informationSTATE VARIANCE CONSTRAINED FUZZY CONTROL VIA OBSERVER-BASED FUZZY CONTROLLERS
Jounal of Maine Science and echnology, Vol. 4, No., pp. 49-57 (6) 49 SAE VARIANCE CONSRAINED FUZZY CONROL VIA OBSERVER-BASED FUZZY CONROLLERS Wen-Je Chang*, Yi-Lin Yeh**, and Yu-eh Meng*** Key wods: takagi-sugeno
More informationA Three-Dimensional Magnetic Force Solution Between Axially-Polarized Permanent-Magnet Cylinders for Different Magnetic Arrangements
Poceedings of the 213 Intenational Confeence on echanics, Fluids, Heat, Elasticity Electomagnetic Fields A Thee-Dimensional agnetic Foce Solution Between Axially-Polaied Pemanent-agnet Cylindes fo Diffeent
More informationThe Chromatic Villainy of Complete Multipartite Graphs
Rocheste Institute of Technology RIT Schola Wos Theses Thesis/Dissetation Collections 8--08 The Chomatic Villainy of Complete Multipatite Gaphs Anna Raleigh an9@it.edu Follow this and additional wos at:
More informationDescription of TEAM Workshop Problem 33.b: Experimental Validation of Electric Local Force Formulations.
1 Desciption of TEAM Wokshop Poblem 33.b: Expeimental Validation of Electic Local Foce Fomulations. Olivie Baé, Pascal Bochet, Membe, IEEE Abstact Unde extenal stesses, the esponse of any mateial is a
More informationME 210 Applied Mathematics for Mechanical Engineers
Tangent and Ac Length of a Cuve The tangent to a cuve C at a point A on it is defined as the limiting position of the staight line L though A and B, as B appoaches A along the cuve as illustated in the
More informationPROBLEM SET #1 SOLUTIONS by Robert A. DiStasio Jr.
POBLM S # SOLUIONS by obet A. DiStasio J. Q. he Bon-Oppenheime appoximation is the standad way of appoximating the gound state of a molecula system. Wite down the conditions that detemine the tonic and
More informationBayesian Congestion Control over a Markovian Network Bandwidth Process
Bayesian Congestion Contol ove a Makovian Netwok Bandwidth Pocess Paisa Mansouifad,, Bhaska Kishnamachai, Taa Javidi Ming Hsieh Depatment of Electical Engineeing, Univesity of Southen Califonia, Los Angeles,
More information