DOCSLIB.ORG

Image Segmentation Combining Markov Random Fields and Dirichlet Processes

Total Page：16

File Type：pdf, Size：1020Kb

Download full-text PDF Read full-text
Image Segmentation Combining Markov Random Fields and Dirichlet Processes ANR meeting Image segmentation combining Markov Random Fields and Dirichlet Processes Jessica SODJO IMS, Groupe Signal Image, Talence Encadrants : A. Giremus, J.-F. Giovannelli, F. Caron, N. Dobigeon Jessica SODJO ANR meeting 1 / 28 ANR meeting Plan 1 Introduction 2 Segmentation using DP models Mixed MRF / DP model Inference : Swendsen-Wang algorithm 3 Hierarchical segmentation with shared classes Principle HDP theory 4 Conclusion and perspective Jessica SODJO ANR meeting 2 / 28 ANR meeting Introduction Segmentation – partition of an image in K homogeneous regions called classes – label the pixels : pixel i $ zi 2 f1;:::; K g Bayesian approach – prior on the distribution of the pixels – all the pixels in a class have the same distribution characterized by a parameter vector Uk – Markov Random Fields (MRF) : exploit the similarity of pixels in the same neighbourhood Constraint : K must be fixed a priori Idea : use the BNP models to directly estimate K Jessica SODJO ANR meeting 3 / 28 ANR meeting Segmentation using DP models Plan 1 Introduction 2 Segmentation using DP models Mixed MRF / DP model Inference : Swendsen-Wang algorithm 3 Hierarchical segmentation with shared classes Principle HDP theory 4 Conclusion and perspective Jessica SODJO ANR meeting 4 / 28 ANR meeting Segmentation using DP models Notations – N is the number of pixels – Y is the observed image – Z = fz1;:::; zN g – Π = fA1;:::; AK g is a partition and m = fm1;:::; mK g with mk = jAk j A1 A2 m1 = 1 m2 = 5 A3 m3 = 6 mK = 4 AK FIGURE: Example of partition Jessica SODJO ANR meeting 5 / 28 ANR meeting Segmentation using DP models Mixed MRF / DP model Markov Random Fields (MRF) – Description of the image by a neighbouring system Considered pixel Neighbours 4-neighbours 8-neighbours FIGURE: Examples of neighbouring system –A clique c is either a singleton either a set of pixels in the same neighbourhood Jessica SODJO ANR meeting 6 / 28 ANR meeting Segmentation using DP models Mixed MRF / DP model Markov Random Fields Let θi 2 fU1;:::; UK g be the parameter vector associated to the i-th pixel MRF , p(θi j θ−i ) = p(θi j θV(i)) where V(i) is the set of neighbours of pixel i Hammersley-Clifford theorem ) Gibbs field ! 1 1 X p(θ) = exp (−Φ(θ)) = exp − Φ (θ ) (1) Z Z c c Φ Φ c with Φc(θc) the local potential and Φ(θ) the global one Limitation: K is assumed to be known Jessica SODJO ANR meeting 7 / 28 ANR meeting Segmentation using DP models Mixed MRF / DP model Potts model The Potts model is a special MRF defined by : 0 1 X M(Π) / exp @ βij 1zi =zj A (2) i$j where – i $ j means that the pixels i and j are neighbours – βij > 0 if i and j are neighbours and βij = 0 otherwise Jessica SODJO ANR meeting 8 / 28 ANR meeting Segmentation using DP models Mixed MRF / DP model The DP model k−1 0 0 Y 0 τk j γ; H ∼ Beta(1; γ) τk = τk (1 − τl ) (3) l=1 where Beta(:) is the Beta distribution P1 Let us write τ ∼ Stick(γ), τ = fτ1; τ2;:::g and k=1 τk = 1 1 X G j γ; H ∼ DP(γ; H) G = τk δUk (4) k=1 with iid Uk j H ∼ H (5) The distribution of the observations is f , defined as : yi j θi ∼ f (: j θi ) and θi j G ∼ G (6) Jessica SODJO ANR meeting 9 / 28 ANR meeting Segmentation using DP models Mixed MRF / DP model The DP model The Chinese Restaurant Process says, K −i −i X m γ θ j θ ∼ k δ + H i −i N − 1 + γ Uk N − 1 + γ k=1 −i – mk is the size of cluster k if we remove pixel i from the partition – K −i is the number of clusters in the image with the i-th pixel removed – Uk is the parameter vector associated to the k-th cluster Limitation : the spatial interactions are not taken into account Jessica SODJO ANR meeting 10 / 28 ANR meeting Segmentation using DP models Mixed MRF / DP model Principle of the segmentation using DP models Define a distribution on the partitions using : – a model that allows that pixels in the same neighbourhood are likely to be in the same cluster (MRF) – DP model to deduce automatically the number of clusters (and if needed their parameters) Jessica SODJO ANR meeting 11 / 28 ANR meeting Segmentation using DP models Mixed MRF / DP model Prior distribution mixing DP and MRF 1 X 1 X p(θ) / exp(− Φi (θi )) exp(− Φc(θc)) ZG ZM i c2C2 | {z } | {z } Ψ(θ) DP model M(θ) MRF model where – C2 means jcj > 2 and j:j is the size. – Φi (:) is defined as : N Z Y Φi (θi ) = − log G(θi ) and ZG = exp(− log G(θi ))dθ1 ::: dθN i=1 N Y ) Ψ(θ) = G(θi ) i=1 P. Orbanz & J. M. Buhmann Nonparametric Bayesian image segmentation, International Journal of Computer Vision, 2007 Jessica SODJO ANR meeting 12 / 28 ANR meeting Segmentation using DP models Mixed MRF / DP model Prior distribution mixing DP and MRF We can deduce : K X −i γ P(θi j θ−i ) / M(θi j θ−i )mk δUk + H (7) ZΦ k=1 Probability of assignment to a new cluster : Z qi0 / f (yi j θ)H(θ)dθ (8) Ωθ Probability of assignment to an existing cluster : −i qik / mk exp(−Φ(Uk j θ−i ))f (yi j Uk ) (9) Parameter update : Y Uk ∼ G0(Uk ) f (yi j Uk ) (10) iji2Ak Jessica SODJO ANR meeting 13 / 28 ANR meeting Segmentation using DP models Inference : Swendsen-Wang algorithm Swendsen-Wang algorithm : principle * Estimation based on the joint posterior p(θ; Z j Y ) * Intractable ) Markov Chain Monte Carlo (MCMC) Problem : very slow convergence Goal : Sample faster the partition of the image – Introduction of a new set of latent variables r such that : p(Π; r) = p(Π)p(r j Π) Q p(r j Π) = p(rij j Π) 1<i<j<N p(rij = 1 j Π) = 1 − exp(βij δij 1zi =zj ) The marginal posterior p(θ; Z j Y ) is unchanged – The links define the "so-called" spin-clusters Jessica SODJO ANR meeting 14 / 28 ANR meeting Segmentation using DP models Inference : Swendsen-Wang algorithm Swendsen-Wang algorithm : principle – Update the labels of the spin-clusters This operation update simultaneously the labels of all the pixels in a spin-cluster FIGURE: Example of label update for spin-clusters Jessica SODJO ANR meeting 15 / 28 ANR meeting Segmentation using DP models Inference : Swendsen-Wang algorithm Swendsen-Wang algorithm : principle – rij ∼ Ber(1 − exp(βij δij 1zi =zj )) with Ber(:) is the Bernouilli distribution Let S = fS1;:::; Spg be the set of spin-clusters. – While removing the spin-cluster Sl , Π = fA−l ;:::; A−l g is the partition obtained while −l 1 K−l removing all pixels in spin-cluster Sl −l −l mk = jAk j Jessica SODJO ANR meeting 16 / 28 ANR meeting Segmentation using DP models Inference : Swendsen-Wang algorithm Swendsen-Wang algorithm : principle For l = 1 : p * The probability to assign pixels in spin-cluster Sl to cluster k is : −l −l −l −l qlk / Ψ(m ;:::; m + jSl j;:::; m )p(ySl j y ) 1 k K−l Ak Q exp(βij (1 − δij )1zi =zj ) f(i;j)ji2Sl ;rij =0g * The probability to assign pixels in spin-cluster Sl to a new cluster is : q = Ψ(m−l ;:::; m−l ; jS j)p(y ) l0 1 K−l l Sl R Q with p(yAk ) = f (yi j Uk )H(Uk )dUk i2Ak Jessica SODJO ANR meeting 17 / 28 ANR meeting Hierarchical segmentation with shared classes Plan 1 Introduction 2 Segmentation using DP models Mixed MRF / DP model Inference : Swendsen-Wang algorithm 3 Hierarchical segmentation with shared classes Principle HDP theory 4 Conclusion and perspective Jessica SODJO ANR meeting 18 / 28 ANR meeting Hierarchical segmentation with shared classes Principle Proposed idea – Different levels of classification can be considered – Coarse categories : urban, sub-urban, forest, etc. – Sub-classes shared between the categories : trees, roads, buildings Taking into account the fact that the classes are shared between different categories can help estimating their parameters and thereby improve the segmentation Jessica SODJO ANR meeting 19 / 28 ANR meeting Hierarchical segmentation with shared classes HDP theory Solution : Hierarchical DP Let J be the number of categories G0 j γ; H ∼ DP(γ; H) Gj j α0; G0 ∼ DP(α0; G0) for j = 1;:::; J ∗ α0 2 R+ G0 is a discrete distribution Discreteness of G0 ) clusters shared among categories Jessica SODJO ANR meeting 20 / 28 ANR meeting Hierarchical segmentation with shared classes HDP theory 1 X G0 = τk δUk (11) k=1 where τ jγ ∼ Stick(γ), τ = fτ1; τ2;:::g and Uk j H ∼ H 1 X Gj = πjk δUk (12) k=1 with πj j α0; τ ∼ DP(α0; τ ) and πj = fπj1; πj2;:::g 'ji j Gj ∼ Gj (13) So, samples of the processes G0 and Gj can be seen as infinite countable mixtures of Dirac measures with respective coefficients τ and πj . Jessica SODJO ANR meeting 21 / 28 ANR meeting Hierarchical segmentation with shared classes HDP theory Principle - Chinese Restaurant Franchise NOTATIONS – J restaurants – Same menu for all restaurants - U1; U2;::: – Tj is the number of tables in restaurant j – θjt is the t-th table of restaurant j – 'ji is the i-th client in restaurant j – njt is the number of clients at a table t – ηjk is the number of tables in restaurant j P which have chosen dish Uk and ηk = k ηjk Jessica SODJO ANR meeting 22 / 28 ANR meeting Hierarchical segmentation with shared classes HDP theory Principle - Chinese Restaurant Franchise Restaurant 1 '11 '12 '14 '13 U 1 U 2 U 2 = = = ..
Load more

Recommended publications
  • Lecturenotes 4 MCMC I – Contents
    Lecturenotes 4 MCMC I – Contents 1. Statistical Physics and Potts Models 2. Sampling and Re-weighting 3. Importance Sampling and Markov Chain Monte Carlo 4. The Metropolis Algorithm 5. The Heatbath Algorithm (Gibbs Sampler) 6. Start and Equilibration 7. Energy Checks 8. Specific Heat 1 Statistical Physics and Potts Model MC simulations of systems described by the Gibbs canonical ensemble aim at calculating estimators of physical observables at a temperature T . In the following we choose units so that β = 1/T and consider the calculation of the expectation value of an observable O. Mathematically all systems on a computer are discrete, because a finite word length has to be used. Hence, K X (k) Ob = Ob(β) = hOi = Z−1 O(k) e−β E (1) k=1 K X (k) where Z = Z(β) = e−β E (2) k=1 is the partition function. The index k = 1,...,K labels all configurations (or microstates) of the system, and E(k) is the (internal) energy of configuration k. To distinguish the configuration index from other indices, it is put in parenthesis. 2 We introduce generalized Potts models in an external magnetic field on d- dimensional hypercubic lattices with periodic boundary conditions. Without being overly complicated, these models are general enough to illustrate the essential features we are interested in. In addition, various subcases of these models are by themselves of physical interest. Generalizations of the algorithmic concepts to other models are straightforward, although technical complications may arise. We define the energy function of the system by (k) (k) (k) −β E = −β E0 + HM (3) where (k) X (k) (k) (k) (k) 2 d N E = −2 J (q , q ) δ(q , q ) + (4) 0 ij i j i j q hiji N 1 for qi = qj (k) X (k) with δ(qi, qj) = and M = 2 δ(1, qi ) .
    [Show full text]
  • Mathematisches Forschungsinstitut Oberwolfach Scaling Limits in Models of Statistical Mechanics
    Mathematisches Forschungsinstitut Oberwolfach Report No. 41/2018 DOI: 10.4171/OWR/2018/41 Scaling Limits in Models of Statistical Mechanics Organised by Dmitry Ioffe, Haifa Gady Kozma, Rehovot Fabio Toninelli, Lyon 9 September – 15 September 2018 Abstract. This conference (part of a long running series) aims to cover the interplay between probability and mathematical statistical mechanics. Specific topics addressed during the 22 talks include: Universality and critical phenomena, disordered models, Gaussian free field (GFF), random planar graphs and unimodular planar maps, reinforced random walks and non-linear σ-models, non-equilibrium dynamics. Less stress is given to topics which have running series of Oberwolfach conferences devoted to them specifically, such as random matrices or integrable models and KPZ universality class. There were 50 participants, including 9 postdocs and graduate students, working in diverse intertwining areas of probability, statistical mechanics and mathematical physics. Subject classification: MSC: 60,82; IMU: 10,13. Introduction by the Organisers This workshop was a sequel to a MFO conference, by the same organizers, which took place in 2015. More broadly, it is a sequel to MFO conferences in 2006, 2009 and 2012, organised by Ken Alexander, Marek Biskup, Remco van der Hofstad and Vladas Sidoravicius. The main focus of the conference remained on probabilistic and analytic methods of non-integrable statistical mechanics. With respect to the previous editions, greater emphasis was put on statistical mechanics models on groups and general graphs, as a lot has happened in this arena recently. The list of 50 participants reflects our attempts to maintain an optimal balance between diverse fields, leading experts and promising young researchers.
    [Show full text]
  • Random and Out-Of-Equilibrium Potts Models Christophe Chatelain
    Random and Out-of-Equilibrium Potts models Christophe Chatelain To cite this version: Christophe Chatelain. Random and Out-of-Equilibrium Potts models. Statistical Mechanics [cond- mat.stat-mech]. Université de Lorraine, 2012. tel-00959733 HAL Id: tel-00959733 https://tel.archives-ouvertes.fr/tel-00959733 Submitted on 15 Mar 2014 HAL is a multi-disciplinary open access L’archive ouverte pluridisciplinaire HAL, est archive for the deposit and dissemination of sci- destinée au dépôt et à la diffusion de documents entific research documents, whether they are pub- scientifiques de niveau recherche, publiés ou non, lished or not. The documents may come from émanant des établissements d’enseignement et de teaching and research institutions in France or recherche français ou étrangers, des laboratoires abroad, or from public or private research centers. publics ou privés. Habilitation `aDiriger des Recherches Mod`eles de Potts d´esordonn´eset hors de l’´equilibre Christophe Chatelain Soutenance publique pr´evue le 17 d´ecembre 2012 Membres du jury : Rapporteurs : Werner Krauth Ecole´ Normale Sup´erieure, Paris Marco Picco Universit´ePierre et Marie Curie, Paris 6 Heiko Rieger Universit´ede Saarbr¨ucken, Allemagne Examinateurs : Dominique Mouhanna Universit´ePierre et Marie Curie, Paris 6 Wolfhard Janke Universit´ede Leipzig, Allemagne Bertrand Berche Universit´ede Lorraine Institut Jean Lamour Facult´edes Sciences - 54500 Vandœuvre-l`es-Nancy Table of contents 1. Random and Out-of-Equililibrium Potts models 4 1.1.Introduction................................. 4 1.2.RandomPottsmodels .......................... 6 1.2.1.ThepurePottsmodel ......................... 6 1.2.1.1. Fortuin-Kasteleyn representation . 7 1.2.1.2. From loop and vertex models to the Coulomb gas .
    [Show full text]
  • Theory of Continuum Percolation I. General Formalism
    Theory of continuum percolation I. General formalism Alon Drory Dipartimento di Fisica, Universit´a La Sapienza Piazzale Aldo Moro 2, Roma 00187, Italy Abstract The theoretical basis of continuum percolation has changed greatly since its beginning as little more than an analogy with lattice systems. Nevertheless, there is yet no comprehensive theory of this field. A ba- sis for such a theory is provided here with the introduction of the Potts fluid, a system of interacting s-state spins which are free to move in the continuum. In the s → 1 limit, the Potts magnetization, suscepti- bility and correlation functions are directly related to the percolation probability, the mean cluster size and the pair-connectedness, respec- tively. Through the Hamiltonian formulation of the Potts fluid, the standard methods of statistical mechanics can therefore be used in the continuum percolation problem. arXiv:cond-mat/9606195v1 26 Jun 1996 1 1 Introduction The theoretical basis of continuum percolation [1] changed drastically during the 1970’s. In their seminal 1970 paper, Scher and Zallen [2] considered continuum percolation as an extension of lattice models. Noting that the critical volume φc is approximately universal, i. e., essentially independent of the lattice structure, they suggested it might be carried over into continuum systems. Such a view of the continuum as the representation of what is lattice- independent sits well with the ideas of the renormalization group where the continuum often represents merely a scale at which the underlying lattice blurs and vanishes. Yet continuum percolation turned out to be much richer than suggested by this conception.
    [Show full text]
  • CFT Interpretation of Merging Multiple SLE Traces
    CFT Interpretation of Merging Multiple SLE Traces Annekathrin Muller-Lohmann¨ ∗ Institut f¨ur theoretische Physik University of Hannover Appelstraße 2 D-30167 Hannover, Germany November 1, 2018 Abstract In this paper we give a physical interpretation of the probability of a Stochastic Lowner¨ Evolution (SLE) trace approaching a marked point in the upper half plane, e. g. on another trace. Our approach is based on the concept of fusion of boundary with bulk or other boundary fields in Conformal Field Theory (CFT). In this context we also state a proposal how the results can be extended to situations that happen not sufficiently far away from the boundary. Contents 1 Introduction 2 1.1 Motivation .............................. 2 1.2 Outline ................................ 3 2 Basic Definitions and Notations 3 2.1 SingleSLE .............................. 3 2.2 MultipleSLE ............................. 4 2.3 BoundaryCFTRevisited ...................... 5 2.4 SLE Martingales and Physical Quantities in BCFT . 7 2.5 FusionandtheOPEinCFT .................... 10 3 ProbabilitiesinBCFTandSLE 11 arXiv:0711.3165v1 [math-ph] 20 Nov 2007 3.1 Partition Functions and Probabilities in BCFT . 11 3.2 The SLE Probability of Intersecting a Disc . 12 4 TheEffectofFusionontheScalingProperties 13 4.1 TheL¨owner EquationDifferential Operators. 13 4.2 ThePurelyBoundaryDifferentialOperators. 14 5 Interpretation of Merging multiple SLE traces 15 5.1 SLE Traces Visiting a Point in the Upper Half Plane . 15 5.2 SLETracesMerging ......................... 17 5.3 TheAngularDependency . 18 ∗[email protected] 6 MultipleSLEasMartingaleWeightedSLE 19 6.1 The Associated Martingale Mt ................... 19 6.2 Probability Interpretation for Mt .................. 20 6.3 Fusion in the Associated Martingale . 21 7 Discussion 22 A Appendix: How to Compute “Fused” Differential Operators 23 1 Introduction 1.1 Motivation Stochastic Lowner¨ Evolutions (SLE) as introduced by O.
    [Show full text]
  • A Novel Approach for Markov Random Field with Intractable Normalising Constant on Large Lattices
    A novel approach for Markov Random Field with intractable normalising constant on large lattices W. Zhu ∗ and Y. Fany February 19, 2018 Abstract The pseudo likelihood method of Besag (1974), has remained a popular method for estimating Markov random field on a very large lattice, despite various documented deficiencies. This is partly because it remains the only computationally tractable method for large lattices. We introduce a novel method to estimate Markov random fields defined on a regular lattice. The method takes advantage of conditional independence structures and recur- sively decomposes a large lattice into smaller sublattices. An approximation is made at each decomposition. Doing so completely avoids the need to com- pute the troublesome normalising constant. The computational complexity arXiv:1601.02410v1 [stat.ME] 11 Jan 2016 is O(N), where N is the the number of pixels in lattice, making it computa- tionally attractive for very large lattices. We show through simulation, that the proposed method performs well, even when compared to the methods using exact likelihoods. Keywords: Markov random field, normalizing constant, conditional indepen- dence, decomposition, Potts model. ∗School of Mathematics and Statistics, University of New South Wales, Sydney 2052 Australia. Communicating Author Wanchuang Zhu: Email [email protected]. ySchool of Mathematics and Statistics, University of New South Wales, Sydney 2052 Australia. Email [email protected]. 1 1 Introduction Markov random field (MRF) models have an important role in modelling spa- tially correlated datasets. They have been used extensively in image and texture analyses ( Nott and Ryden´ 1999, Hurn et al. 2003), image segmentation (Pal and Pal 1993, Van Leemput et al.
    [Show full text]
  • A Review on Statistical Inference Methods for Discrete Markov Random Fields Julien Stoehr, Richard Everitt, Matthew T
    A review on statistical inference methods for discrete Markov random fields Julien Stoehr, Richard Everitt, Matthew T. Moores To cite this version: Julien Stoehr, Richard Everitt, Matthew T. Moores. A review on statistical inference methods for discrete Markov random fields. 2017. hal-01462078v2 HAL Id: hal-01462078 https://hal.archives-ouvertes.fr/hal-01462078v2 Preprint submitted on 11 Apr 2017 HAL is a multi-disciplinary open access L’archive ouverte pluridisciplinaire HAL, est archive for the deposit and dissemination of sci- destinée au dépôt et à la diffusion de documents entific research documents, whether they are pub- scientifiques de niveau recherche, publiés ou non, lished or not. The documents may come from émanant des établissements d’enseignement et de teaching and research institutions in France or recherche français ou étrangers, des laboratoires abroad, or from public or private research centers. publics ou privés. A review on statistical inference methods for discrete Markov random fields Julien Stoehr1 1School of Mathematical Sciences & Insight Centre for Data Analytics, University College Dublin, Ireland Abstract Developing satisfactory methodology for the analysis of Markov random field is a very challenging task. Indeed, due to the Markovian dependence structure, the normalizing constant of the fields cannot be computed using standard analytical or numerical methods. This forms a central issue for any statistical approach as the likelihood is an integral part of the procedure. Furthermore, such unobserved fields cannot be integrated out and the likelihood evaluation becomes a doubly intractable problem. This report gives an overview of some of the methods used in the literature to analyse such observed or unobserved random fields.
    [Show full text]
  • Critical Gaussian Multiplicative Chaos: Convergence of The
    The Annals of Probability 2014, Vol. 42, No. 5, 1769–1808 DOI: 10.1214/13-AOP890 c Institute of Mathematical Statistics, 2014 CRITICAL GAUSSIAN MULTIPLICATIVE CHAOS: CONVERGENCE OF THE DERIVATIVE MARTINGALE By Bertrand Duplantier1, Remi´ Rhodes2, Scott Sheffield3 and Vincent Vargas4 Institut de Physique Th´eorique, Universit´eParis-Dauphine, Massachusetts Institute of Technology and Universit´eParis-Dauphine In this paper, we study Gaussian multiplicative chaos in the criti- cal case. We show that the so-called derivative martingale, introduced in the context of branching Brownian motions and branching random walks, converges almost surely (in all dimensions) to a random mea- sure with full support. We also show that the limiting measure has no atom. In connection with the derivative martingale, we write explicit conjectures about the glassy phase of log-correlated Gaussian poten- tials and the relation with the asymptotic expansion of the maximum of log-correlated Gaussian random variables. 1. Introduction. 1.1. Overview. In the 1980s, Kahane [45] developed a continuous param- eter theory of multifractal random measures, called Gaussian multiplicative chaos; this theory emerged from the need to define rigorously the limit log- normal model introduced by Mandelbrot [59] in the context of turbulence. His efforts were followed by several authors [3, 7, 11, 35, 67–69] coming up with various generalizations at different scales. This family of random fields has found many applications in various fields of science, especially in turbulence and in mathematical finance. Recently, the authors in [30] arXiv:1206.1671v3 [math.PR] 2 Sep 2014 Received July 2012; revised September 2013. 1Supported in part by Grant ANR-08-BLAN-0311-CSD5 and by the MISTI MIT- France Seed Fund.
    [Show full text]
  • The Potts Model and Tutte Polynomial, and Associated Connections Between Statistical Mechanics and Graph Theory
    The Potts Model and Tutte Polynomial, and Associated Connections Between Statistical Mechanics and Graph Theory Robert Shrock C. N. Yang Institute for Theoretical Physics, Stony Brook University Lecture 1 at Indiana University - Purdue University, Indianapolis (IUPUI) Workshop on Connections Between Complex Dynamics, Statistical Physics, and Limiting Spectra of Self-similar Group Actions, Aug. 2016 Outline • Introduction • Potts model in statistical mechanics and equivalence of Potts partition function with Tutte polynomial in graph theory; special cases • Some easy cases • Calculational methods and some simple examples • Partition functions zeros and their accumulation sets as n → ∞ • Chromatic polynomials and ground state entropy of Potts antiferromagnet • Historical note on graph coloring • Conclusions Introduction Some Results from Statistical Physics Statistical physics deals with properties of many-body systems. Description of such systems uses a specified form for the interaction between the dynamical variables. An example is magnetic systems, where the dynamical variables are spins located at sites of a regular lattice Λ; these interact with each other by an energy function called a Hamiltonian . H Simple example: a model with integer-valued (effectively classical) spins σ = 1 at i ± sites i on a lattice Λ, with Hamiltonian = J σ σ H σ H − I i j − i Xeij Xi where JI is the spin-spin interaction constant, eij refers to a bond on Λ joining sites i and j, and H is a possible external magnetic field. This is the Ising model. Let T denote the temperature and define β = 1/(kBT ), where k = 1.38 10 23 J/K=0.862 10 4 eV/K is the Boltzmann constant.
    [Show full text]
  • Markov Chain Monte Carlo Sampling
    Markov Chain Monte Carlo Sampling Bachelors Thesis Project Bachelors of Technology in Computer Science and Engineering by Saurabh Garg (140070003) under the guidance of Prof.Suyash Awate Department of Computer Science and Engineering Indian Institute of Technology Bombay November, 2017 Abstract Markov chain Monte Carlo (MCMC) methods are a class of algorithms for sampling from a probability distribution based on constructing a Markov chain that has the desired distribution as its equilibrium distribution. The state of the chain after a number of steps is then used as a sample of the desired distribution. The quality of the sample improves as a function of the number of steps. Though a Markov chain gives us a neat solution, the problem with using a Markov chain is that it converges to its stationary distribution on when its simulated for a long time. To tackle this problem, the concept of perfect sampling was introduced which can tell if the the chain we are simulating has converged or not. In this report, we would be getting an idea about Markov chains first and then go through various MCMC sampling methods in existence. We then see how perfect sampling fits into some of these methods and extend it over posteriors. We extend perfect sampling algorithms to sample from general posteriors using bounding chains. We also see, how we can use the posterior sampling methods for MRI Image Segmentation and uncertainty estimation. Declaration I declare that this written submission represents my ideas in my own words and where other’s ideas or words have been included, I have adequately cited and referenced the original sources.
    [Show full text]
  • Sharp Phase Transition for the Random-Cluster and Potts Models Via Decision Trees
    Sharp phase transition for the random-cluster and Potts models via decision trees Hugo Duminil-Copin∗†, Aran Raoufi†, Vincent Tassion‡ December 23, 2018 Abstract We prove an inequality on decision trees on monotonic measures which generalizes the OSSS inequality on product spaces. As an application, we use this inequality to prove a number of new results on lattice spin models and their random-cluster representations. More precisely, we prove that • For the Potts model on transitive graphs, correlations decay exponentially fast for β < βc. • For the random-cluster model with cluster weight q ≥ 1 on transitive graphs, correlations decay exponentially fast in the subcritical regime and the cluster-density satisfies the mean- field lower bound in the supercritical regime. • For the random-cluster models with cluster weight q ≥ 1 on planar quasi-transitive graphs G, ∗ pc(G)pc(G ) ∗ = q: (1 − pc(G))(1 − pc(G )) As a special case, we obtain the value of the critical point for the square, triangular and hexagonal lattices (this provides a short proof of the result of [BD12]). These results have many applications for the understanding of the subcritical (respectively dis- ordered) phase of all these models. The techniques developed in this paper have potential to be extended to a wide class of models including the Ashkin-Teller model, continuum percolation mod- els such as Voronoi percolation and Boolean percolation, super-level sets of massive Gaussian Free Field, and random-cluster and Potts model with infinite range interactions. 1 Introduction 1.1 OSSS inequality for monotonic measures In theoretical computer science, determining the computational complexity of tasks is a very difficult problem (think of P against NP).
    [Show full text]
  • MRF), Potts Model, Wireless Sensor Networks (Wsns
    American Journal of Intelligent Systems 2012, 2(7): 157-162 DOI: 10.5923/j.ajis.20120207.01 Adaptive Scheduling in Wireless Sensor Networks Based on Potts Model Afsoon Assa, Majid Vafaei Jahan* Department of Computer Engineering, Islamic Azad University, Mashhad Branch, Mashhad, Iran Abstract Recent advances in wireless sensor network's technology have developed using this technology in various fields. For the importance and intense application of these networks, many efforts and researchers have been done to confront challenges. In this paper, to conquer one of the most important challenges of these networks, that is the limitation in energy resources, an adaptive scheduling algorithm based in Potts model was applied. According to this model, for every element within a system, q different states are considered. Each element converts its state to a new state or remains in that state due to its current state and its adjacent neighbours and with the effect of environment. In this project, this model was used in wireless sensor networks such that each sensor node is considered as an element of sensor network system and three active, inactive and standby states are defined for that node, which it selects one of these three states according to its current state and neighbours and environment affect and adapts its activity on the environment in a way that the least energy to be consumed. By comparing this algorithm and similar algorithm (with two states), it is observed that in identical conditions, Potts model with three states represent better results and more lifetimes for network. Keywords Ising Model, Markov Random Field (MRF), Potts Model, Wireless Sensor Networks (Wsns) and unpredictable.
    [Show full text]