UNSUPERVISED NON STATIONARY IMAGE SEGMENTATION USING TRIPLET MARKOV CHAINS. Pierre Lanchantin and Wojciech Pieczynski
|
|
- Stephany Francis
- 5 years ago
- Views:
Transcription
1 Advaced Cocepts for Itelliget Visio Sstems ACVIS 0 Aug. -Sept. Brussels Belgium 00 SPERVISED O STATIOARY IMAGE SEGMETATIO SIG TRIPLET MARKOV CHAIS Pierre Lachati ad Wojciech Pieczski Pierre.Lachati@it-evr.fr GET/IT Départemet CITI CRS MR 7 9 rue Charles Fourier 9000 Evr Frace ABSTRACT This work deals with the usupervised Baesia hidde Markov chai restoratio eteded to the o statioar case. supervised restoratio based o Epectatio- Maimizatio EM or Stochastic EM SEM estimates cosiderig the Hidde Markov Chai HMC model is quite efficiet whe the hidde chai is statioar. However whe the latter is ot statioar the usupervised restoratio results ca be poor due to a bad match betwee the real ad estimated models. I this paper we preset a more appropriate model for o statioar HMC via recet Triplet Markov Chais TMC model. sig TMC we show that the classical restoratio results ca be sigificatl improved i the case of o statioar data. The latter improvemet is performed i a usupervised wa usig a SEM parameter estimatio method. Some applicatio eamples to usupervised image segmetatio are also provided.. ITRODCTIO The hidde Markov chai HMC model is widel used for various problems icludig sigal ad image processig ecoomical predictio ad health scieces. I such model X = X... X is a stochastic process modelig a uobservable or hidde discrete sigal each X takes its values i a fiite set { } Ω = ω...ω K ad Y = Y... Y is a stochastic process modelig the observatios each Y takes its values i the set of real umbers R. The problem is the to estimate the hidde discrete sigal X = from the observed sigal Y =. The liks betwee X ad Y are the modeled b the followig joit distributio : p = p p... p p... p Such a model is called hidde Markov chai with idepedet oise because the hidde process X is a Markov chai ad the radom variables Y... Y are idepedet coditioall o X. Hereafter it will be deoted b HMC-I. It allows oe to recover the hidde data X = from the observed data Y = usig differet Baesia classificatio techiques like Maimum A Posteriori MAP or Maimal Posterior Mode MPM [-]. These restoratio methods use the distributio of the hidde process coditioal to the observatios which is called its posterior distributio. Whe p is based o ukow parameters θ Θ the latter ca be estimated from Y = cosiderig the statioarit of the hidde process b methods like Epectatio-Maimizatio EM [ ] or Stochastic Epectatio-Maimizatio SEM [ ]. Thus whe the hidde process is a statioar Markov chai parameters are well estimated ad the restoratio based o the estimated parameters gives good results. However whe this is ot the case the estimatio ecessaril gives wrog results which ca impl poor restoratio of X. I this work we propose to model the o statioarit b itroducig a auiliar process goverig the regime switchig of the hidde process. The itroductio of a auiliar process to model the o statioarit has alread bee treated before. I [] the authors itroduce auiliar data to modif the trasitio probabilities of the Markov hidde process. I [] the author preset a model based o the superpositio of two Markov chais. A o-homogeeous discrete process is described b a set of trasitio matrices. At each the choice of the active matri is the govered b a hidde homogeeous Markov chai. evertheless our approach is quite differet : we propose a model of o statioar hidde Markov chai b usig the Triplet Markov Chai TMC model [78]. It is based o the three followig poits : i Whe X is o statioar we ca itroduce a auiliar stochastic process whose aim is to model the regime switchig of X ; ii We directl assume the markoviait ad the statioarit of the pairwise process V = X. The the distributios P X ad P are ot ecessaril Markovia which allows oe to deal with differet kids of stochastic processes; iii The distributio of Y coditioal o V = X is such that the triplet process T = X Y is a Markov chai. Although the margial distributio P of T is X Y
2 Advaced Cocepts for Itelliget Visio Sstems ACVIS 0 Aug. -Sept. Brussels Belgium 00 ot ecessaril a Markov distributio it ca be used to perform Baesia restoratios T beig a particular case of Triplet Markov chai [7]. We provide differet simulatio studies showig that such a itroductio of a auiliar process ca improve the results obtaied with the classical usupervised Baesia restoratio. We provide a complete derivatio of the Forward-Backward MPM ad SEM algorithms for the ew model ad give computable versios of these oes. The paper is orgaized i the followig wa : The et sectio is devoted to a brief descriptio of the Triplet Markov Chai model. The particular model used ad the related processig method is described i sectio three ad differet simulatio results described i sectio four.. TRIPLET MARKOV CHAIS Let X = X ad Y = Y be two stochastic processes. X is hidde each X takes its values i a fiite set Ω = { } ω...ω K ad Y is observed each Y takes its values i the set of real umbers R. The problem is the to estimate X = from Y =. Defiitio. The model cosidered is called a Triplet Markov Chai if there eists a stochastic process = with λ M takig its values i a set Λ = { λ... } each such T = X Y = X Y is a Markov chai. that The idea of TMC is to cosider the distributio of Z = X Y which models the iteractios betwee the observed ad the searched process as a margial distributio of a Markov chai T = X Y. The distributio s p which are used i MPM restoratio are u Λ u the computable. I fact let V = X. As T is Markovia the process V Y is a Pairwise Markov Chai PMC ad we ca formulate ad calculate the distributio of V Y usig Forward-Backward recursios [9]. This meas that the distributios p = p u = p v also are computable. Fiall although the distributio of X Y is ot ecessaril a Markov oe the margial distributios p are computable. which i particular eables the use of the Baesia MPM restoratio method. We said above that Z = X Y i ot ecessaril a Markov chai. More precisel the followig result specifies locall o what the greater geeralit of TMC with respect to PMC cosists [8] : Λ Propositio. Let T = X Y be a TMC verifig : a p t t + does ot deped o ; ad b p t t = + p t + t. The the three followig coditios: i Z = X Y is a Markov chai ; iifor each p u z z = p u z ; iiifor each p u z = p u z are equivalet. The Pairwise Markov Chai PMC [90] model i which Z is assumed to be a Markov chai is the a particular case of the TMC obtaied b takig Λ = Ω ad = X. Furthermore accordig to the Propositio. above TMC is strictl more geeral tha PMC the latter beig strictl more geeral tha classical HMC i that there eist Markov chais T such that Z are ot Markov chais.. THE PARTICLAR TMC SED AD RELATED PROCESSIG METHOD.. TMC used model Let X be a o statioar hidde process X. Let us itroduce =... each takes its values i a fiite set Λ = { λ... λ M } goverig the regime switchig of X. The pairwise process V = X is assumed to be a Markov chai. Further we assume that Y...Y are idepedet coditioall o X ad p u = p. The T = V Y is a particular TMC i which V is a Markov chai ad which will be called TMC i the followig... MPM restoratio i TMC Accordig to the geeral Baesia theor the MPM restoratio is give b s... = ˆ... ˆ with ˆ MPM ˆ = arg ma p Ω Thus we eed to calculate p. As T is a TMC the process V Y is a PMC ad we ca write the distributio of V Y as p v = α v β v with. α v = p v... Forward probabilities ad v = p... + v β Backward probabilities.
3 Advaced Cocepts for Itelliget Visio Sstems ACVIS 0 Aug. -Sept. Brussels Belgium 00 These probabilities ca the be recursivel calculated for b α v = p v p ; v = + α v p v+ v p + + v Ω Λ v = α β ; β v β v p v v p = v Ω Λ + The margial posterior distributios of the hidde state ca be calculated b p v α v β v ad the trasitio of the posterior Markov distributio p v are calculated b p v v... + p v v p β v Havig calculated p v we ca the calculate p = p v ad to perform MPM. u Λ.. Learig TMC We assume that p are Gaussia. For K classes { } Ω = ω...ω K we have to estimate K meas µ µ K ad K variaces σ σ K of the K Gaussia desities p = p = K. Further the distributio of the statioar Markov chai V = X K M K M is give b parameters p ij = p v = i v = which is a probabilit o [ Λ] j Ω. The SEM method we use rus as follows : i cosider a iitial values p µ σ K Λ 0 i j ad k K ; ii for each q * : -simulate q V = v accordig to v which is possible due to ; q+ q+ q+ q+ θ = p µ σ with -calculate ij k k q+ q+ p ij = [ v = i v + = j ] k = q+ + = µ k q [ = k ] k = = [ = k ] ij k k θ = for p based o θ q µ = = [ = k ] = [ = k ] σ 7. SPERVISED IMAGE SEGMETATIO SIG TMC I this sectio we compare the HMC-I ad the proposed TMC models i the field of image segmetatio. The HMC-I model has alread bee successfull applied to this problem : to do so the bi-dimesioal set of piels is trasformed ito a moo-dimesioal sequece through the Hilbert-Peao sca [0]. Parameters are the estimated with SEM algorithm ad the MPM is applied to perform the segmetatio. We preset two series of eperimets. I the first oe two sthetic hidde o statioar images are cosidered. We give differet usupervised restoratio results cosiderig the HMC ad the TMC models ad show that the latter is more appropriate whe the hidde process presets differet regimes. I the secod oe we compare the two models cosiderig a real image usupervised segmetatio... Sthetic images We cosider two HMC X Y verifig with { } Ω = ω ω ad = 8 8. I the two HMC the hidde processes are o-statioar i differet wa. We defie three trasitio matrices : M = M = ad M = ad two auiliar processes =... ad =... each ad Λ = λ λ λ. takes its values i the fiite set { } * is defie b :... / = λ = λ = λ. / +... / ad / +... * is a Markov chai defie as follow : i the distributio of is /// ii the trasitio matri of is give b M = I the first eperimet the Markov chai X = X... is o statioar i the followig X wa. i the distributio of X is ii M p is the trasitio matri i with p = λ X whe p
4 Advaced Cocepts for Itelliget Visio Sstems ACVIS 0 Aug. -Sept. Brussels Belgium 00 I the secod eperimet the Markov chai X is obtai i the same wa b replacig b. We will deote = ω = ω = ω λ λ ω λ ω λ ω λ = = =. λ The processes X X are the coverted via a Hilbert-Peao sca ito a bi-dimesioal set as described i [] ad their realizatios are preseted i Figure. X ˆ = ˆ TMC τ = % X ˆ = ˆ TMC τ =.% = u X = Y = X ˆ = ˆ HMC τ = % = u X = Y = X ˆ = ˆ HMC τ = 9%. ˆ = uˆ TMC ˆ = uˆ TMC Figure : Differet images ad their usupervised HMC ad TMC based restoratios. I the two eperimets a realizatio X = is simulated ad Y = is sampled accordig to p p where p = is Gaussia with mea ad variace ad p = is Gaussia with mea ad variace which gives Y = ad Y = i Figure. The realizatio X = is the estimated b the Baesia MPM method from Y = i two differet was. The first restoratio is obtaied usig the parameters estimated with the SEM algorithm ad cosiderig that X Y is a HMC-I the hidde chai is assumed statioar which gives X ˆ = ˆ HMC τ = % ad X ˆ = ˆ HMC τ = 9% i Figure τ is the error ratio. The secod restoratio is obtaied usig the parameters estimated with the SEM algorithm assumig that T = X Y is the TMC cosidered above which gives X ˆ = ˆ TMC τ = % ad X ˆ = ˆ TMC τ =.% i Figure. We also show i Figure the estimates ˆ = uˆ ad ˆ = uˆ of the of the auiliar processes = u ad = u. We clearl see how the use of TMC ca improve the results obtaied with the classical HMC-I model. Furthermore differet parameters are well estimated Table ad which show a good behavior of the SEM cosidered.
5 Advaced Cocepts for Itelliget Visio Sstems ACVIS 0 Aug. -Sept. Brussels Belgium 00 Model µ σ µ σ p HMC TMC p HMC TMC Table : estimated desities parameters. The best model is the oe with the lowest BIC. Results are preseted i Table. a e e e b Table : estimated a p v + v ad b p cosiderig the TMC model. a e- 0 9e b Table : estimated a p v + v ad b p cosiderig the TMC model... Real Image Our eample here is a size satellite image of Toko which is preseted i Figure. We restore the image cosiderig differet umber of real ad auiliar states. Comparisos betwee models are carried out usig the Baes Iformatio Criterio BIC []. This Criterio is defied as BIC = LL + p log where LL is the log-likelihood of the model p its umber of idepedet parameters ad the umber of data. We do ot take ito accout parameters estimated to be zeros accordig to the covetio established i []. v 0 v 0 Figure : Image Toko O Figure we preset the segmetatio result b MPM ad SEM parameters estimatio based o the statioar HMC model with classes. O Figure we preset the segmetatio result b MPM ad SEM parameters estimatio based o the TMC model with classes ad states auiliar process. umber of real states umber of auiliar states LLumber of parameters BIC Table : Comparisos betwee models. Accordig to the BIC values we ca sa that for each umber of states studied the TMC model with the largest umber of auiliar states is more appropriate. Furthermore we ca see b comparig segmetatio results o Figure ad Figure that the segmetatio result cosiderig the TMC model preset fier features.
6 Advaced Cocepts for Itelliget Visio Sstems ACVIS 0 Aug. -Sept. Brussels Belgium 00 precisel the o statioar prior distributio of the hidde Markov chai was modeled b a auiliar process goverig the switchig of the trasitio matri. Baesia segmetatio techiques are the redered applicable ad differet eperimets show its iterest i simulated ad real images.. REFERECES Figure : MPM segmetatio after SEM parameters estimatio based o the statioar HMC mode with real states. BIC=80. Figure : MPM segmetatio after SEM parameters estimatio based o the TMC model with real states ad auiliar states. BIC=0.. COCLSIOS AD PERSPECTIVES I this paper we have provided a usupervised method for restorig o statioar hidde Markov chais with potetial applicatios to usupervised image segmetatio. The mai cotributio was to tackle the lack of statioarit usig the Triplet Markov Chai model. More. L. E. Baum T. Petrie G. Soules. Weiss. A maimizatio techique occurig i the statistical aalsis of probabilistic fuctios of Markov chais A. Math. Statistic. pp G. D. Fora The Viterbi algorithm Proceedigs of the IEEE Vol. o. pp G. J. McLachla ad T. Krisha EM Algorithm ad Etesios Wile G. Celeu ad J. Diebolt The SEM algorithm: a probabilistic teacher derived from the EM algorithm for the miture problem Comput. Statist. Quarter. pp J. P. Hugues P. Guttorp ad S. P. Charles A ohomogeeous hidde Markov model for precipitatio occurrece Appl.stat vol. 8 pp A. Berchtold The double chai Markov model Commu. Statist. Theor Methods W. Pieczski Chaîes de Markov Triplet Triplet Markov Chais Comptes Redus de l Académie des Scieces Mathématiques Paris Ser. I pp W. Pieczski Triplet Markov Chais ad Theor of Evidece Submitted to Iteratioal Joural of Approimate Reasoig September W. Pieczski Pairwise Markov chais IEEE Tras. o PAMI Vol. o pp S. Derrode ad W. Pieczski SAR image segmetatio usig geeralized Pairwise Markov Chais SPIE s Iteratioal Smposium o Remote Sesig September -7 Crete Greece 00.. S. Derrode ad W. Pieczski Sigal ad image restoratio usig Pairwise Markov Chais IEEE Tras. o Sigal Processig scheduled for September Giordaa ad W. Pieczski Estimatio of Geeralized Multisesor Hidde Markov Chais ad supervised Image Segmetatio IEEE Tras. o PAMI Vol. 9 o. pp R. W. Katz O some criteria for estimatig the order of a Markov chai Techometrics Y. M. Bishop S.E. Fieberg P. W Hollad Discrete Multivariate Aalsis MIT Press Cambridge 97.
EECS564 Estimation, Filtering, and Detection Hwk 2 Solns. Winter p θ (z) = (2θz + 1 θ), 0 z 1
EECS564 Estimatio, Filterig, ad Detectio Hwk 2 Sols. Witer 25 4. Let Z be a sigle observatio havig desity fuctio where. p (z) = (2z + ), z (a) Assumig that is a oradom parameter, fid ad plot the maximum
More informationExpectation-Maximization Algorithm.
Expectatio-Maximizatio Algorithm. Petr Pošík Czech Techical Uiversity i Prague Faculty of Electrical Egieerig Dept. of Cyberetics MLE 2 Likelihood.........................................................................................................
More informationFast Exact Filtering in Generalized Conditionally Observed Markov Switching Models with Copulas
Fast Exact Filterig i Geeralized Coditioally Observed Markov Switchig Models with Copulas Fei Zheg, Stéphae Derrode, Wojciech Pieczyski To cite this versio: Fei Zheg, Stéphae Derrode, Wojciech Pieczyski.
More informationINF Introduction to classifiction Anne Solberg Based on Chapter 2 ( ) in Duda and Hart: Pattern Classification
INF 4300 90 Itroductio to classifictio Ae Solberg ae@ifiuioo Based o Chapter -6 i Duda ad Hart: atter Classificatio 90 INF 4300 Madator proect Mai task: classificatio You must implemet a classificatio
More informationRAINFALL PREDICTION BY WAVELET DECOMPOSITION
RAIFALL PREDICTIO BY WAVELET DECOMPOSITIO A. W. JAYAWARDEA Departmet of Civil Egieerig, The Uiversit of Hog Kog, Hog Kog, Chia P. C. XU Academ of Mathematics ad Sstem Scieces, Chiese Academ of Scieces,
More informationCS434a/541a: Pattern Recognition Prof. Olga Veksler. Lecture 5
CS434a/54a: Patter Recogitio Prof. Olga Veksler Lecture 5 Today Itroductio to parameter estimatio Two methods for parameter estimatio Maimum Likelihood Estimatio Bayesia Estimatio Itroducto Bayesia Decisio
More informationInformation-based Feature Selection
Iformatio-based Feature Selectio Farza Faria, Abbas Kazeroui, Afshi Babveyh Email: {faria,abbask,afshib}@staford.edu 1 Itroductio Feature selectio is a topic of great iterest i applicatios dealig with
More informationNew Ratio Estimators Using Correlation Coefficient
New atio Estimators Usig Correlatio Coefficiet Cem Kadilar ad Hula Cigi Hacettepe Uiversit, Departmet of tatistics, Betepe, 06800, Akara, Turke. e-mails : kadilar@hacettepe.edu.tr ; hcigi@hacettepe.edu.tr
More informationA Family of Unbiased Estimators of Population Mean Using an Auxiliary Variable
Advaces i Computatioal Scieces ad Techolog ISSN 0973-6107 Volume 10, Number 1 (017 pp. 19-137 Research Idia Publicatios http://www.ripublicatio.com A Famil of Ubiased Estimators of Populatio Mea Usig a
More informationRandom Variables, Sampling and Estimation
Chapter 1 Radom Variables, Samplig ad Estimatio 1.1 Itroductio This chapter will cover the most importat basic statistical theory you eed i order to uderstad the ecoometric material that will be comig
More informationClustering. CM226: Machine Learning for Bioinformatics. Fall Sriram Sankararaman Acknowledgments: Fei Sha, Ameet Talwalkar.
Clusterig CM226: Machie Learig for Bioiformatics. Fall 216 Sriram Sakararama Ackowledgmets: Fei Sha, Ameet Talwalkar Clusterig 1 / 42 Admiistratio HW 1 due o Moday. Email/post o CCLE if you have questios.
More informationAn Introduction to Randomized Algorithms
A Itroductio to Radomized Algorithms The focus of this lecture is to study a radomized algorithm for quick sort, aalyze it usig probabilistic recurrece relatios, ad also provide more geeral tools for aalysis
More informationA Note on Effi cient Conditional Simulation of Gaussian Distributions. April 2010
A Note o Effi ciet Coditioal Simulatio of Gaussia Distributios A D D C S S, U B C, V, BC, C April 2010 A Cosider a multivariate Gaussia radom vector which ca be partitioed ito observed ad uobserved compoetswe
More informationECE 8527: Introduction to Machine Learning and Pattern Recognition Midterm # 1. Vaishali Amin Fall, 2015
ECE 8527: Itroductio to Machie Learig ad Patter Recogitio Midterm # 1 Vaishali Ami Fall, 2015 tue39624@temple.edu Problem No. 1: Cosider a two-class discrete distributio problem: ω 1 :{[0,0], [2,0], [2,2],
More informationThe Sampling Distribution of the Maximum. Likelihood Estimators for the Parameters of. Beta-Binomial Distribution
Iteratioal Mathematical Forum, Vol. 8, 2013, o. 26, 1263-1277 HIKARI Ltd, www.m-hikari.com http://d.doi.org/10.12988/imf.2013.3475 The Samplig Distributio of the Maimum Likelihood Estimators for the Parameters
More informationGeneralized Semi- Markov Processes (GSMP)
Geeralized Semi- Markov Processes (GSMP) Summary Some Defiitios Markov ad Semi-Markov Processes The Poisso Process Properties of the Poisso Process Iterarrival times Memoryless property ad the residual
More informationLainiotis filter implementation. via Chandrasekhar type algorithm
Joural of Computatios & Modellig, vol.1, o.1, 2011, 115-130 ISSN: 1792-7625 prit, 1792-8850 olie Iteratioal Scietific Press, 2011 Laiiotis filter implemetatio via Chadrasehar type algorithm Nicholas Assimais
More informationTHE KALMAN FILTER RAUL ROJAS
THE KALMAN FILTER RAUL ROJAS Abstract. This paper provides a getle itroductio to the Kalma filter, a umerical method that ca be used for sesor fusio or for calculatio of trajectories. First, we cosider
More informationDouble Stage Shrinkage Estimator of Two Parameters. Generalized Exponential Distribution
Iteratioal Mathematical Forum, Vol., 3, o. 3, 3-53 HIKARI Ltd, www.m-hikari.com http://dx.doi.org/.9/imf.3.335 Double Stage Shrikage Estimator of Two Parameters Geeralized Expoetial Distributio Alaa M.
More informationModeling and Estimation of a Bivariate Pareto Distribution using the Principle of Maximum Entropy
Sri Laka Joural of Applied Statistics, Vol (5-3) Modelig ad Estimatio of a Bivariate Pareto Distributio usig the Priciple of Maximum Etropy Jagathath Krisha K.M. * Ecoomics Research Divisio, CSIR-Cetral
More informationDimension-free PAC-Bayesian bounds for the estimation of the mean of a random vector
Dimesio-free PAC-Bayesia bouds for the estimatio of the mea of a radom vector Olivier Catoi CREST CNRS UMR 9194 Uiversité Paris Saclay olivier.catoi@esae.fr Ilaria Giulii Laboratoire de Probabilités et
More informationBayesian and E- Bayesian Method of Estimation of Parameter of Rayleigh Distribution- A Bayesian Approach under Linex Loss Function
Iteratioal Joural of Statistics ad Systems ISSN 973-2675 Volume 12, Number 4 (217), pp. 791-796 Research Idia Publicatios http://www.ripublicatio.com Bayesia ad E- Bayesia Method of Estimatio of Parameter
More informationEntropy Rates and Asymptotic Equipartition
Chapter 29 Etropy Rates ad Asymptotic Equipartitio Sectio 29. itroduces the etropy rate the asymptotic etropy per time-step of a stochastic process ad shows that it is well-defied; ad similarly for iformatio,
More informationThe Poisson Process *
OpeStax-CNX module: m11255 1 The Poisso Process * Do Johso This work is produced by OpeStax-CNX ad licesed uder the Creative Commos Attributio Licese 1.0 Some sigals have o waveform. Cosider the measuremet
More informationOn forward improvement iteration for stopping problems
O forward improvemet iteratio for stoppig problems Mathematical Istitute, Uiversity of Kiel, Ludewig-Mey-Str. 4, D-24098 Kiel, Germay irle@math.ui-iel.de Albrecht Irle Abstract. We cosider the optimal
More informationEstimation for Complete Data
Estimatio for Complete Data complete data: there is o loss of iformatio durig study. complete idividual complete data= grouped data A complete idividual data is the oe i which the complete iformatio of
More informationQuantile regression with multilayer perceptrons.
Quatile regressio with multilayer perceptros. S.-F. Dimby ad J. Rykiewicz Uiversite Paris 1 - SAMM 90 Rue de Tolbiac, 75013 Paris - Frace Abstract. We cosider oliear quatile regressio ivolvig multilayer
More informationInformation Theory Tutorial Communication over Channels with memory. Chi Zhang Department of Electrical Engineering University of Notre Dame
Iformatio Theory Tutorial Commuicatio over Chaels with memory Chi Zhag Departmet of Electrical Egieerig Uiversity of Notre Dame Abstract A geeral capacity formula C = sup I(; Y ), which is correct for
More informationf(x i ; ) L(x; p) = i=1 To estimate the value of that maximizes L or equivalently ln L we will set =0, for i =1, 2,...,m p x i (1 p) 1 x i i=1
Parameter Estimatio Samples from a probability distributio F () are: [,,..., ] T.Theprobabilitydistributio has a parameter vector [,,..., m ] T. Estimator: Statistic used to estimate ukow. Estimate: Observed
More informationMATH 320: Probability and Statistics 9. Estimation and Testing of Parameters. Readings: Pruim, Chapter 4
MATH 30: Probability ad Statistics 9. Estimatio ad Testig of Parameters Estimatio ad Testig of Parameters We have bee dealig situatios i which we have full kowledge of the distributio of a radom variable.
More informationCSE 527, Additional notes on MLE & EM
CSE 57 Lecture Notes: MLE & EM CSE 57, Additioal otes o MLE & EM Based o earlier otes by C. Grat & M. Narasimha Itroductio Last lecture we bega a examiatio of model based clusterig. This lecture will be
More informationVector Quantization: a Limiting Case of EM
. Itroductio & defiitios Assume that you are give a data set X = { x j }, j { 2,,, }, of d -dimesioal vectors. The vector quatizatio (VQ) problem requires that we fid a set of prototype vectors Z = { z
More informationJanuary 25, 2017 INTRODUCTION TO MATHEMATICAL STATISTICS
Jauary 25, 207 INTRODUCTION TO MATHEMATICAL STATISTICS Abstract. A basic itroductio to statistics assumig kowledge of probability theory.. Probability I a typical udergraduate problem i probability, we
More informationMarkov Decision Processes
Markov Decisio Processes Defiitios; Statioary policies; Value improvemet algorithm, Policy improvemet algorithm, ad liear programmig for discouted cost ad average cost criteria. Markov Decisio Processes
More informationBasis for simulation techniques
Basis for simulatio techiques M. Veeraraghava, March 7, 004 Estimatio is based o a collectio of experimetal outcomes, x, x,, x, where each experimetal outcome is a value of a radom variable. x i. Defiitios
More informationMOMENT-METHOD ESTIMATION BASED ON CENSORED SAMPLE
Vol. 8 o. Joural of Systems Sciece ad Complexity Apr., 5 MOMET-METHOD ESTIMATIO BASED O CESORED SAMPLE I Zhogxi Departmet of Mathematics, East Chia Uiversity of Sciece ad Techology, Shaghai 37, Chia. Email:
More informationBIRKHOFF ERGODIC THEOREM
BIRKHOFF ERGODIC THEOREM Abstract. We will give a proof of the poitwise ergodic theorem, which was first proved by Birkhoff. May improvemets have bee made sice Birkhoff s orgial proof. The versio we give
More informationChandrasekhar Type Algorithms. for the Riccati Equation of Lainiotis Filter
Cotemporary Egieerig Scieces, Vol. 3, 00, o. 4, 9-00 Chadrasekhar ype Algorithms for the Riccati Equatio of Laiiotis Filter Nicholas Assimakis Departmet of Electroics echological Educatioal Istitute of
More informationConfidence interval for the two-parameter exponentiated Gumbel distribution based on record values
Iteratioal Joural of Applied Operatioal Research Vol. 4 No. 1 pp. 61-68 Witer 2014 Joural homepage: www.ijorlu.ir Cofidece iterval for the two-parameter expoetiated Gumbel distributio based o record values
More informationECE 901 Lecture 12: Complexity Regularization and the Squared Loss
ECE 90 Lecture : Complexity Regularizatio ad the Squared Loss R. Nowak 5/7/009 I the previous lectures we made use of the Cheroff/Hoeffdig bouds for our aalysis of classifier errors. Hoeffdig s iequality
More informationSequential Monte Carlo Methods - A Review. Arnaud Doucet. Engineering Department, Cambridge University, UK
Sequetial Mote Carlo Methods - A Review Araud Doucet Egieerig Departmet, Cambridge Uiversity, UK http://www-sigproc.eg.cam.ac.uk/ ad2/araud doucet.html ad2@eg.cam.ac.uk Istitut Heri Poicaré - Paris - 2
More informationThe Expectation-Maximization (EM) Algorithm
The Expectatio-Maximizatio (EM) Algorithm Readig Assigmets T. Mitchell, Machie Learig, McGraw-Hill, 997 (sectio 6.2, hard copy). S. Gog et al. Dyamic Visio: From Images to Face Recogitio, Imperial College
More informationComputational Tutorial of Steepest Descent Method and Its Implementation in Digital Image Processing
Computatioal utorial of Steepest Descet Method ad Its Implemetatio i Digital Image Processig Vorapoj Pataavijit Departmet of Electrical ad Electroic Egieerig, Faculty of Egieerig Assumptio Uiversity, Bagkok,
More informationThere is no straightforward approach for choosing the warmup period l.
B. Maddah INDE 504 Discrete-Evet Simulatio Output Aalysis () Statistical Aalysis for Steady-State Parameters I a otermiatig simulatio, the iterest is i estimatig the log ru steady state measures of performace.
More informationRegression and generalization
Regressio ad geeralizatio CE-717: Machie Learig Sharif Uiversity of Techology M. Soleymai Fall 2016 Curve fittig: probabilistic perspective Describig ucertaity over value of target variable as a probability
More informationProbabilistic Classifiers Using Nearest Neighbor Balls. Climate Change Workshop, Malta, March, 2009
Probabilistic Classifiers Usig Nearest Neighbor Balls Climate Chage Worshop Malta March 2009 Bo Raeby & Ju Yu Cetre of Biostochastics Swedish Uiversity of Agricultural Scieces Bacgroud Climate chage implies
More informationChapter 12 EM algorithms The Expectation-Maximization (EM) algorithm is a maximum likelihood method for models that have hidden variables eg. Gaussian
Chapter 2 EM algorithms The Expectatio-Maximizatio (EM) algorithm is a maximum likelihood method for models that have hidde variables eg. Gaussia Mixture Models (GMMs), Liear Dyamic Systems (LDSs) ad Hidde
More informationPattern Classification
Patter Classificatio All materials i these slides were tae from Patter Classificatio (d ed) by R. O. Duda, P. E. Hart ad D. G. Stor, Joh Wiley & Sos, 000 with the permissio of the authors ad the publisher
More informationFirst Year Quantitative Comp Exam Spring, Part I - 203A. f X (x) = 0 otherwise
First Year Quatitative Comp Exam Sprig, 2012 Istructio: There are three parts. Aswer every questio i every part. Questio I-1 Part I - 203A A radom variable X is distributed with the margial desity: >
More informationTopics Machine learning: lecture 2. Review: the learning problem. Hypotheses and estimation. Estimation criterion cont d. Estimation criterion
.87 Machie learig: lecture Tommi S. Jaakkola MIT CSAIL tommi@csail.mit.edu Topics The learig problem hypothesis class, estimatio algorithm loss ad estimatio criterio samplig, empirical ad epected losses
More informationLaw of the sum of Bernoulli random variables
Law of the sum of Beroulli radom variables Nicolas Chevallier Uiversité de Haute Alsace, 4, rue des frères Lumière 68093 Mulhouse icolas.chevallier@uha.fr December 006 Abstract Let be the set of all possible
More information1.010 Uncertainty in Engineering Fall 2008
MIT OpeCourseWare http://ocw.mit.edu.00 Ucertaity i Egieerig Fall 2008 For iformatio about citig these materials or our Terms of Use, visit: http://ocw.mit.edu.terms. .00 - Brief Notes # 9 Poit ad Iterval
More informationProperties and Hypothesis Testing
Chapter 3 Properties ad Hypothesis Testig 3.1 Types of data The regressio techiques developed i previous chapters ca be applied to three differet kids of data. 1. Cross-sectioal data. 2. Time series data.
More informationStatistical Inference (Chapter 10) Statistical inference = learn about a population based on the information provided by a sample.
Statistical Iferece (Chapter 10) Statistical iferece = lear about a populatio based o the iformatio provided by a sample. Populatio: The set of all values of a radom variable X of iterest. Characterized
More informationIntroducing a Novel Bivariate Generalized Skew-Symmetric Normal Distribution
Joural of mathematics ad computer Sciece 7 (03) 66-7 Article history: Received April 03 Accepted May 03 Available olie Jue 03 Itroducig a Novel Bivariate Geeralized Skew-Symmetric Normal Distributio Behrouz
More informationMachine Learning Regression I Hamid R. Rabiee [Slides are based on Bishop Book] Spring
Machie Learig Regressio I Hamid R. Rabiee [Slides are based o Bishop Book] Sprig 015 http://ce.sharif.edu/courses/93-94//ce717-1 Liear Regressio Liear regressio: ivolves a respose variable ad a sigle predictor
More informationEstimation of the Mean and the ACVF
Chapter 5 Estimatio of the Mea ad the ACVF A statioary process {X t } is characterized by its mea ad its autocovariace fuctio γ ), ad so by the autocorrelatio fuctio ρ ) I this chapter we preset the estimators
More informationMaximum Likelihood Estimation
Chapter 9 Maximum Likelihood Estimatio 9.1 The Likelihood Fuctio The maximum likelihood estimator is the most widely used estimatio method. This chapter discusses the most importat cocepts behid maximum
More informationEmpirical Process Theory and Oracle Inequalities
Stat 928: Statistical Learig Theory Lecture: 10 Empirical Process Theory ad Oracle Iequalities Istructor: Sham Kakade 1 Risk vs Risk See Lecture 0 for a discussio o termiology. 2 The Uio Boud / Boferoi
More informationMA Advanced Econometrics: Properties of Least Squares Estimators
MA Advaced Ecoometrics: Properties of Least Squares Estimators Karl Whela School of Ecoomics, UCD February 5, 20 Karl Whela UCD Least Squares Estimators February 5, 20 / 5 Part I Least Squares: Some Fiite-Sample
More informationComparison of Minimum Initial Capital with Investment and Non-investment Discrete Time Surplus Processes
The 22 d Aual Meetig i Mathematics (AMM 207) Departmet of Mathematics, Faculty of Sciece Chiag Mai Uiversity, Chiag Mai, Thailad Compariso of Miimum Iitial Capital with Ivestmet ad -ivestmet Discrete Time
More informationJoint Probability Distributions and Random Samples. Jointly Distributed Random Variables. Chapter { }
UCLA STAT A Applied Probability & Statistics for Egieers Istructor: Ivo Diov, Asst. Prof. I Statistics ad Neurology Teachig Assistat: Neda Farziia, UCLA Statistics Uiversity of Califoria, Los Ageles, Sprig
More informationThe Sample Variance Formula: A Detailed Study of an Old Controversy
The Sample Variace Formula: A Detailed Study of a Old Cotroversy Ky M. Vu PhD. AuLac Techologies Ic. c 00 Email: kymvu@aulactechologies.com Abstract The two biased ad ubiased formulae for the sample variace
More informationPattern Classification
Patter Classificatio All materials i these slides were tae from Patter Classificatio (d ed) by R. O. Duda, P. E. Hart ad D. G. Stor, Joh Wiley & Sos, 000 with the permissio of the authors ad the publisher
More informationThe DOA Estimation of Multiple Signals based on Weighting MUSIC Algorithm
, pp.10-106 http://dx.doi.org/10.1457/astl.016.137.19 The DOA Estimatio of ultiple Sigals based o Weightig USIC Algorithm Chagga Shu a, Yumi Liu State Key Laboratory of IPOC, Beijig Uiversity of Posts
More informationR. van Zyl 1, A.J. van der Merwe 2. Quintiles International, University of the Free State
Bayesia Cotrol Charts for the Two-parameter Expoetial Distributio if the Locatio Parameter Ca Take o Ay Value Betwee Mius Iity ad Plus Iity R. va Zyl, A.J. va der Merwe 2 Quitiles Iteratioal, ruaavz@gmail.com
More informationBayesian Methods: Introduction to Multi-parameter Models
Bayesia Methods: Itroductio to Multi-parameter Models Parameter: θ = ( θ, θ) Give Likelihood p(y θ) ad prior p(θ ), the posterior p proportioal to p(y θ) x p(θ ) Margial posterior ( θ, θ y) is Iterested
More informationMathematical Modeling of Optimum 3 Step Stress Accelerated Life Testing for Generalized Pareto Distribution
America Joural of Theoretical ad Applied Statistics 05; 4(: 6-69 Published olie May 8, 05 (http://www.sciecepublishiggroup.com/j/ajtas doi: 0.648/j.ajtas.05040. ISSN: 6-8999 (Prit; ISSN: 6-9006 (Olie Mathematical
More informationLecture 15: Learning Theory: Concentration Inequalities
STAT 425: Itroductio to Noparametric Statistics Witer 208 Lecture 5: Learig Theory: Cocetratio Iequalities Istructor: Ye-Chi Che 5. Itroductio Recall that i the lecture o classificatio, we have see that
More informationProbability and MLE.
10-701 Probability ad MLE http://www.cs.cmu.edu/~pradeepr/701 (brief) itro to probability Basic otatios Radom variable - referrig to a elemet / evet whose status is ukow: A = it will rai tomorrow Domai
More informationStatistical Pattern Recognition
Statistical Patter Recogitio Classificatio: No-Parametric Modelig Hamid R. Rabiee Jafar Muhammadi Sprig 2014 http://ce.sharif.edu/courses/92-93/2/ce725-2/ Ageda Parametric Modelig No-Parametric Modelig
More informationA statistical method to determine sample size to estimate characteristic value of soil parameters
A statistical method to determie sample size to estimate characteristic value of soil parameters Y. Hojo, B. Setiawa 2 ad M. Suzuki 3 Abstract Sample size is a importat factor to be cosidered i determiig
More informationThe random version of Dvoretzky s theorem in l n
The radom versio of Dvoretzky s theorem i l Gideo Schechtma Abstract We show that with high probability a sectio of the l ball of dimesio k cε log c > 0 a uiversal costat) is ε close to a multiple of the
More informationImproved Class of Ratio -Cum- Product Estimators of Finite Population Mean in two Phase Sampling
Global Joural of Sciece Frotier Research: F Mathematics ad Decisio Scieces Volume 4 Issue 2 Versio.0 Year 204 Type : Double Blid Peer Reviewed Iteratioal Research Joural Publisher: Global Jourals Ic. (USA
More informationTable 12.1: Contingency table. Feature b. 1 N 11 N 12 N 1b 2 N 21 N 22 N 2b. ... a N a1 N a2 N ab
Sectio 12 Tests of idepedece ad homogeeity I this lecture we will cosider a situatio whe our observatios are classified by two differet features ad we would like to test if these features are idepedet
More informationSome limit properties for a hidden inhomogeneous Markov chain
Dog et al. Joural of Iequalities ad Applicatios (208) 208:292 https://doi.org/0.86/s3660-08-884-7 R E S E A R C H Ope Access Some it properties for a hidde ihomogeeous Markov chai Yu Dog *, Fag-qig Dig
More informationADVANCED SOFTWARE ENGINEERING
ADVANCED SOFTWARE ENGINEERING COMP 3705 Exercise Usage-based Testig ad Reliability Versio 1.0-040406 Departmet of Computer Ssciece Sada Narayaappa, Aeliese Adrews Versio 1.1-050405 Departmet of Commuicatio
More informationx iu i E(x u) 0. In order to obtain a consistent estimator of β, we find the instrumental variable z which satisfies E(z u) = 0. z iu i E(z u) = 0.
27 However, β MM is icosistet whe E(x u) 0, i.e., β MM = (X X) X y = β + (X X) X u = β + ( X X ) ( X u ) \ β. Note as follows: X u = x iu i E(x u) 0. I order to obtai a cosistet estimator of β, we fid
More informationA General Family of Estimators for Estimating Population Variance Using Known Value of Some Population Parameter(s)
Rajesh Sigh, Pakaj Chauha, Nirmala Sawa School of Statistics, DAVV, Idore (M.P.), Idia Floreti Smaradache Uiversity of New Meico, USA A Geeral Family of Estimators for Estimatig Populatio Variace Usig
More informationAlgorithms for Clustering
CR2: Statistical Learig & Applicatios Algorithms for Clusterig Lecturer: J. Salmo Scribe: A. Alcolei Settig: give a data set X R p where is the umber of observatio ad p is the umber of features, we wat
More informationElementary manipulations of probabilities
Elemetary maipulatios of probabilities Set probability of multi-valued r.v. {=Odd} = +3+5 = /6+/6+/6 = ½ X X,, X i j X i j Multi-variat distributio: Joit probability: X true true X X,, X X i j i j X X
More informationLecture 27: Optimal Estimators and Functional Delta Method
Stat210B: Theoretical Statistics Lecture Date: April 19, 2007 Lecture 27: Optimal Estimators ad Fuctioal Delta Method Lecturer: Michael I. Jorda Scribe: Guilherme V. Rocha 1 Achievig Optimal Estimators
More informationGoodness-Of-Fit For The Generalized Exponential Distribution. Abstract
Goodess-Of-Fit For The Geeralized Expoetial Distributio By Amal S. Hassa stitute of Statistical Studies & Research Cairo Uiversity Abstract Recetly a ew distributio called geeralized expoetial or expoetiated
More informationInternational Journal of Mathematical Archive-5(7), 2014, Available online through ISSN
Iteratioal Joural of Mathematical Archive-5(7), 214, 11-117 Available olie through www.ijma.ifo ISSN 2229 546 USING SQUARED-LOG ERROR LOSS FUNCTION TO ESTIMATE THE SHAPE PARAMETER AND THE RELIABILITY FUNCTION
More informationConvergence of random variables. (telegram style notes) P.J.C. Spreij
Covergece of radom variables (telegram style otes).j.c. Spreij this versio: September 6, 2005 Itroductio As we kow, radom variables are by defiitio measurable fuctios o some uderlyig measurable space
More informationRainy and Dry Days as a Stochastic Process (Albaha City)
IOSR Joural of Mathematics (IOSR-JM) e-issn: 2278-5728, p-issn: 239-765X. Volume, Issue Ver. III (Ja - Feb. 25), PP 86-9 www.iosrjourals.org Raiy ad Dry Days as a Stochastic Process (Albaha City) Dr. Najeeb
More informationA Question. Output Analysis. Example. What Are We Doing Wrong? Result from throwing a die. Let X be the random variable
A Questio Output Aalysis Let X be the radom variable Result from throwig a die 5.. Questio: What is E (X? Would you throw just oce ad take the result as your aswer? Itroductio to Simulatio WS/ - L 7 /
More informationBayesian inference for Parameter and Reliability function of Inverse Rayleigh Distribution Under Modified Squared Error Loss Function
Australia Joural of Basic ad Applied Scieces, (6) November 26, Pages: 24-248 AUSTRALIAN JOURNAL OF BASIC AND APPLIED SCIENCES ISSN:99-878 EISSN: 239-844 Joural home page: www.ajbasweb.com Bayesia iferece
More informationCHAPTER 5 SOME MINIMAX AND SADDLE POINT THEOREMS
CHAPTR 5 SOM MINIMA AND SADDL POINT THORMS 5. INTRODUCTION Fied poit theorems provide importat tools i game theory which are used to prove the equilibrium ad eistece theorems. For istace, the fied poit
More informationEconomics 241B Relation to Method of Moments and Maximum Likelihood OLSE as a Maximum Likelihood Estimator
Ecoomics 24B Relatio to Method of Momets ad Maximum Likelihood OLSE as a Maximum Likelihood Estimator Uder Assumptio 5 we have speci ed the distributio of the error, so we ca estimate the model parameters
More informationKolmogorov-Smirnov type Tests for Local Gaussianity in High-Frequency Data
Proceedigs 59th ISI World Statistics Cogress, 5-30 August 013, Hog Kog (Sessio STS046) p.09 Kolmogorov-Smirov type Tests for Local Gaussiaity i High-Frequecy Data George Tauche, Duke Uiversity Viktor Todorov,
More informationt distribution [34] : used to test a mean against an hypothesized value (H 0 : µ = µ 0 ) or the difference
EXST30 Backgroud material Page From the textbook The Statistical Sleuth Mea [0]: I your text the word mea deotes a populatio mea (µ) while the work average deotes a sample average ( ). Variace [0]: The
More informationGoodness-of-Fit Tests and Categorical Data Analysis (Devore Chapter Fourteen)
Goodess-of-Fit Tests ad Categorical Data Aalysis (Devore Chapter Fourtee) MATH-252-01: Probability ad Statistics II Sprig 2019 Cotets 1 Chi-Squared Tests with Kow Probabilities 1 1.1 Chi-Squared Testig................
More informationStatistical Analysis on Uncertainty for Autocorrelated Measurements and its Applications to Key Comparisons
Statistical Aalysis o Ucertaity for Autocorrelated Measuremets ad its Applicatios to Key Comparisos Nie Fa Zhag Natioal Istitute of Stadards ad Techology Gaithersburg, MD 0899, USA Outlies. Itroductio.
More informationA proposed discrete distribution for the statistical modeling of
It. Statistical Ist.: Proc. 58th World Statistical Cogress, 0, Dubli (Sessio CPS047) p.5059 A proposed discrete distributio for the statistical modelig of Likert data Kidd, Marti Cetre for Statistical
More informationEksamen 2006 H Utsatt SENSORVEILEDNING. Problem 1. Settet består av 9 delspørsmål som alle anbefales å telle likt. Svar er gitt i <<.. >>.
Eco 43 Eksame 6 H Utsatt SENSORVEILEDNING Settet består av 9 delspørsmål som alle abefales å telle likt. Svar er gitt i . Problem a. Let the radom variable (rv.) X be expoetially distributed with
More informationA Note on Box-Cox Quantile Regression Estimation of the Parameters of the Generalized Pareto Distribution
A Note o Box-Cox Quatile Regressio Estimatio of the Parameters of the Geeralized Pareto Distributio JM va Zyl Abstract: Makig use of the quatile equatio, Box-Cox regressio ad Laplace distributed disturbaces,
More informationA New Multivariate Markov Chain Model with Applications to Sales Demand Forecasting
Iteratioal Coferece o Idustrial Egieerig ad Systems Maagemet IESM 2007 May 30 - Jue 2 BEIJING - CHINA A New Multivariate Markov Chai Model with Applicatios to Sales Demad Forecastig Wai-Ki CHING a, Li-Mi
More informationThe r-generalized Fibonacci Numbers and Polynomial Coefficients
It. J. Cotemp. Math. Scieces, Vol. 3, 2008, o. 24, 1157-1163 The r-geeralized Fiboacci Numbers ad Polyomial Coefficiets Matthias Schork Camillo-Sitte-Weg 25 60488 Frakfurt, Germay mschork@member.ams.org,
More informationControl chart for number of customers in the system of M [X] / M / 1 Queueing system
Iteratioal Joural of Iovative Research i Sciece, Egieerig ad Techology (A ISO 3297: 07 Certified Orgaiatio) Cotrol chart for umber of customers i the system of M [X] / M / Queueig system T.Poogodi, Dr.
More information