Mixing times via super-fast coupling
|
|
- Austen Anderson
- 6 years ago
- Views:
Transcription
1 s via super-fast coupling Yevgeniy Kovchegov Department of Mathematics Oregon State University
2 Outline 1 About 2 Definition Coupling Result.
3 About Outline 1 About 2 Definition Coupling Result.
4 About Coauthor Paper. This presentation is based on joint work with R.Burton.
5 About History Diaconis and Shahshahani (early 80 s) The mixing time for shuffling a deck of n cards by random transpositions is of order O(n log(n)) with cut-off asymptotics at 1 2 n log(n). Method used: relatively rarified mathematical residential district of representation theory.
6 About The Problem Open problem (Y.Peres) Provide a coupling proof of O(n log(n)) mixing rate. Continuous time: < a, b > has rate 2 n 2, i.e. < a, b > and < b, a > happen with rate 1 n 2 each. Transposition < a, a > has rate 1 n 2.
7 Outline 1 About 2 Definition Coupling Result.
8 Definition Definition of Mixing Time Total variation distance: µ ν TV := 1 µ(x) ν(x) 2 : process {X t } with stationary distribution π. If X t ν t = ν 0 P t, { t mix := inf t : ν t π TV 1 } 4, all ν 0 x
9 Coupling via coupling X t ν 0 P t and Y t µ 0 P t. If T coupling is the coupling time for (X t, Y t ), then ν 0 P t µ 0 P t TV P[T coupling > t] Goal: bound P[T coupling > t] for all µ 0. Here max x ν 0 P t xp t TV ν 0 P t π TV as π(x) ν 0 P t xp t TV ν 0 P t π TV x by convexity of f (x) = ν 0 P t xp t TV.
10 Coupling Notations Notations <, > - transpositions < ( a, ) > - transposition initiated by card a At - the coupled process B t ( ) At, - transpositions in B t
11 Coupling Coupling #1. (Aldous and Fill) a, i : moves card a to location i in both processes, A t and B t. Mixing: order O(n 2 ) instead of O(n log n); E[T coupling ] n d=2 n 2 ( ) π 2 d n 2 Problem: slows down significantly when the number of discrepancies is small enough, Coupling #2. (equivalent to prev.): applying transposition a, i if a is not coupled, applying transposition a, b if a is coupled Coupling #2 can be improved to match O(n log n) Transpositions a, b called label-to-label.
12 Coupling Group invariance X m is a Markov Chain on a discrete group G. The chain is group invariant if for all γ, α S n. In other words, dist(γx m+1 X m = α) = dist(x m+1 X m = γ 1 α) label-to-label a, b can be ignored. Reason: don t change cycle structure. The situation is invariant under label-to-label transpositions. If label-to-label a, b is applied before label-to-location b, i. We do not relabel, and do not reselect i.
13 Vocabulary. association map - hidden association between positions/locations in the top process and positions/locations in the bottom process that will be used to establish the rates for the coupled process
14 Two discrepancies (d = 2) at d 1 and d 2 : A t : b 9 a 8 a B t : a 9 b 8 a d 2 d 1 i 1 Label-to-location coupling: E[T coupling ] = n2 4 too large.
15 Jump a, i 1 of a to random location i 1 at exponential time t 1 : From to A t : b 9 a 8 a B t : a 9 b 8 a d 2 d 1 i 1 A t : b 9 a1 8 a 2... B t : a1 9 b 8 a 2... d 2 d 1 i 1
16 Different way of saying the same: Start with A t : b 9 a1 8 a 2... B t : a1 9 b 8 a 2... d 2 /i 1 i 1 /d 1 d 1 /d 2 where at time t 1 the locations relabel according to d 1 /d 2 i 1 i 1 /d 1 d 1. d 2 /i 1 d 2
17 Different way of saying the same: Jump a, i 1 at time t 1 exponential ( 1 n). From to A t : b 9 a1 8 a 2... B t : a1 9 b 8 a 2... d 2 /i 1 i 1 /d 1 d 1 /d 2 A t : b 9 a1 8 a 2... B t : a1 9 b 8 a 2... d 2 d 1 i 1
18 The Association Map. The following association map will determine jumps of a1. A t : b 9 a1 8 a 2... B t : a1 9 b 8 a 2... Card a1 will jump to position i 2 on the assoc. map at time t 2, even if t 2 < t 1.
19 A t : b 9 a1 8 a 2... B t : a1 9 b 8 a 2... d2 d1 i1 Now i 2 i 1 and ( t 2 exponential (1 1/n) 1 ) n a1, i 1 = a1, a is label-to-label, we can skip. If i 1 = d 1 or d 2, discrepancies cancel at t 1 ; if i2 = d 1 or d 2, discrepancies cancel on the assoc. map at t 2. If t 1 < t 2, assoc. map real picture at t 1, we create one more assoc. map.
20 Case t 2 < t 1, and i2 = d 2. On association map: Start with A t : b 9 a1 8 a 2... B t : a1 9 b 8 a 2... d 2 /i 1 i 1 /d 1 d 1 /d 2 At time t 2 : A t : a1 9 b 8 a 2... B t : a1 9 b 8 a 2... d 2 /i 1 i 1 /d 1 d 1 /d 2
21 Case t 2 < t 1, and i2 = d 2. On association map: At time t 2 : A t : a1 9 b 8 a 2... B t : a1 9 b 8 a 2... d 2 /i 1 i 1 /d 1 d 1 /d 2 At time t 1 : A t : a1 9 b 8 a 2... B t : a1 9 b 8 a 2... d 2 d 1 i 1
22 Case t 2 < t 1, and i2 = d 2. Same evolution, original association: Start with A t : b 9 a 8 a B t : a 9 b 8 a d 2 d 1 i 1 At time t 2 : A t : a1 9 a 8 b 2... B t : a 9 b 8 a d 2 d 1 i 1
23 Case t 2 < t 1, and i2 = d 2. Same evolution, original association: At time t 2 : A t : a1 9 a 8 b 2... B t : a 9 b 8 a d 2 d 1 i 1 At time t 1 : A t : a1 9 b 8 a 2... B t : a1 9 b 8 a 2... d 2 d 1 i 1
24 Faster coupling. The new coupling time for two discrepancies: E[T coupling ] n2 8
25 Chain of association maps. a1, i 2 occurs at t 2 A t :... a2 6 b 9 a1 8 a 2... B t :... a2 6 a1 9 b 8 a 2... i 2 d2 d1 i1 d 1 is i 1/d 1 before t 1, and d 1 after t 1 ; d 2 is d 2/i 1 before t 1, and d 2 after t 1 ; i 1 is d 1/d 2 before t 1, and i 1 after t 1.
26 Chain of association maps. New association map: A t :... a1 6 b 9 a2 8 a 2... B t :... a1 6 a2 9 b 8 a 2... d1 /d 2 d2 /i 2 i 2 /d1 i1 where at t 2, d 1 /d 2 i 2 i 2 /d 1 d 1 d 2 /i 2 d 2.
27 Chain of association maps. a2 will do label-to-location jump w.r.t. the following assoc. map d d i A t :... a1 6 b 9 a2 8 a 2... B t :... a1 6 a2 9 b 8 a 2... i2 d2 d1 i1 1 is i 2/d1 before t 2, and d1 after t 2; 2 is d 2 /i 2 before t 2, and d2 after t 2; 2 is d 1 /d 2 before t 2, and i 2 after t 2. a2, i 3 occurs at t 3 exponential ( (1 2/n) 1 n )
28 Chain of association maps. And so on, creating a chain of k = εn association maps. In case of d = 2 discrepancies, the average time of discrepancy cancelation on one of association maps is E[T 2 ] = n 2 4(k + 1) n 4ε.
29 Result. Result. General d: E[T d ] = n 2 2(k + 1)d n 2εd. Coupling time (all discrepancies): [ 1 E[T coupling ] 2ε + κ (1 κ)(κ ε) for any 0 < ε < κ < 1. ] n log n
The cutoff phenomenon for random walk on random directed graphs
The cutoff phenomenon for random walk on random directed graphs Justin Salez Joint work with C. Bordenave and P. Caputo Outline of the talk Outline of the talk 1. The cutoff phenomenon for Markov chains
More informationA note on adiabatic theorem for Markov chains and adiabatic quantum computation. Yevgeniy Kovchegov Oregon State University
A note on adiabatic theorem for Markov chains and adiabatic quantum computation Yevgeniy Kovchegov Oregon State University Introduction Max Born and Vladimir Fock in 1928: a physical system remains in
More informationOn Markov Chain Monte Carlo
MCMC 0 On Markov Chain Monte Carlo Yevgeniy Kovchegov Oregon State University MCMC 1 Metropolis-Hastings algorithm. Goal: simulating an Ω-valued random variable distributed according to a given probability
More informationA note on adiabatic theorem for Markov chains
Yevgeniy Kovchegov Abstract We state and prove a version of an adiabatic theorem for Markov chains using well known facts about mixing times. We extend the result to the case of continuous time Markov
More information215 Problem 1. (a) Define the total variation distance µ ν tv for probability distributions µ, ν on a finite set S. Show that
15 Problem 1. (a) Define the total variation distance µ ν tv for probability distributions µ, ν on a finite set S. Show that µ ν tv = (1/) x S µ(x) ν(x) = x S(µ(x) ν(x)) + where a + = max(a, 0). Show that
More informationMixing time for a random walk on a ring
Mixing time for a random walk on a ring Stephen Connor Joint work with Michael Bate Paris, September 2013 Introduction Let X be a discrete time Markov chain on a finite state space S, with transition matrix
More informationOberwolfach workshop on Combinatorics and Probability
Apr 2009 Oberwolfach workshop on Combinatorics and Probability 1 Describes a sharp transition in the convergence of finite ergodic Markov chains to stationarity. Steady convergence it takes a while to
More informationFlip dynamics on canonical cut and project tilings
Flip dynamics on canonical cut and project tilings Thomas Fernique CNRS & Univ. Paris 13 M2 Pavages ENS Lyon November 5, 2015 Outline 1 Random tilings 2 Random sampling 3 Mixing time 4 Slow cooling Outline
More informationThe coupling method - Simons Counting Complexity Bootcamp, 2016
The coupling method - Simons Counting Complexity Bootcamp, 2016 Nayantara Bhatnagar (University of Delaware) Ivona Bezáková (Rochester Institute of Technology) January 26, 2016 Techniques for bounding
More informationMARKOV CHAINS AND HIDDEN MARKOV MODELS
MARKOV CHAINS AND HIDDEN MARKOV MODELS MERYL SEAH Abstract. This is an expository paper outlining the basics of Markov chains. We start the paper by explaining what a finite Markov chain is. Then we describe
More informationPath Coupling and Aggregate Path Coupling
Path Coupling and Aggregate Path Coupling 0 Path Coupling and Aggregate Path Coupling Yevgeniy Kovchegov Oregon State University (joint work with Peter T. Otto from Willamette University) Path Coupling
More informationFAST INITIAL CONDITIONS FOR GLAUBER DYNAMICS
FAST INITIAL CONDITIONS FOR GLAUBER DYNAMICS EYAL LUBETZKY AND ALLAN SLY Abstract. In the study of Markov chain mixing times, analysis has centered on the performance from a worst-case starting state.
More informationEdge Flip Chain for Unbiased Dyadic Tilings
Sarah Cannon, Georgia Tech,, David A. Levin, U. Oregon Alexandre Stauffer, U. Bath, March 30, 2018 Dyadic Tilings A dyadic tiling is a tiling on [0,1] 2 by n = 2 k dyadic rectangles, rectangles which are
More informationCharacterization of cutoff for reversible Markov chains
Characterization of cutoff for reversible Markov chains Yuval Peres Joint work with Riddhi Basu and Jonathan Hermon 23 Feb 2015 Joint work with Riddhi Basu and Jonathan Hermon Characterization of cutoff
More informationIntroduction to Markov Chains and Riffle Shuffling
Introduction to Markov Chains and Riffle Shuffling Nina Kuklisova Math REU 202 University of Chicago September 27, 202 Abstract In this paper, we introduce Markov Chains and their basic properties, and
More informationL n = l n (π n ) = length of a longest increasing subsequence of π n.
Longest increasing subsequences π n : permutation of 1,2,...,n. L n = l n (π n ) = length of a longest increasing subsequence of π n. Example: π n = (π n (1),..., π n (n)) = (7, 2, 8, 1, 3, 4, 10, 6, 9,
More information88 CONTINUOUS MARKOV CHAINS
88 CONTINUOUS MARKOV CHAINS 3.4. birth-death. Continuous birth-death Markov chains are very similar to countable Markov chains. One new concept is explosion which means that an infinite number of state
More informationA Proof of Aldous Spectral Gap Conjecture. Thomas M. Liggett, UCLA. Stirring Processes
A Proof of Aldous Spectral Gap Conjecture Thomas M. Liggett, UCLA Stirring Processes Given a graph G = (V, E), associate with each edge e E a Poisson process Π e with rate c e 0. Labels are put on the
More informationConfidence intervals for the mixing time of a reversible Markov chain from a single sample path
Confidence intervals for the mixing time of a reversible Markov chain from a single sample path Daniel Hsu Aryeh Kontorovich Csaba Szepesvári Columbia University, Ben-Gurion University, University of Alberta
More information3. The Voter Model. David Aldous. June 20, 2012
3. The Voter Model David Aldous June 20, 2012 We now move on to the voter model, which (compared to the averaging model) has a more substantial literature in the finite setting, so what s written here
More informationLecture 28: April 26
CS271 Randomness & Computation Spring 2018 Instructor: Alistair Sinclair Lecture 28: April 26 Disclaimer: These notes have not been subjected to the usual scrutiny accorded to formal publications. They
More informationLecture 2: September 8
CS294 Markov Chain Monte Carlo: Foundations & Applications Fall 2009 Lecture 2: September 8 Lecturer: Prof. Alistair Sinclair Scribes: Anand Bhaskar and Anindya De Disclaimer: These notes have not been
More informationConvergence Rate of Markov Chains
Convergence Rate of Markov Chains Will Perkins April 16, 2013 Convergence Last class we saw that if X n is an irreducible, aperiodic, positive recurrent Markov chain, then there exists a stationary distribution
More informationCharacterization of cutoff for reversible Markov chains
Characterization of cutoff for reversible Markov chains Yuval Peres Joint work with Riddhi Basu and Jonathan Hermon 3 December 2014 Joint work with Riddhi Basu and Jonathan Hermon Characterization of cutoff
More informationIntroduction to Machine Learning CMU-10701
Introduction to Machine Learning CMU-10701 Markov Chain Monte Carlo Methods Barnabás Póczos & Aarti Singh Contents Markov Chain Monte Carlo Methods Goal & Motivation Sampling Rejection Importance Markov
More informationTOTAL VARIATION CUTOFF IN BIRTH-AND-DEATH CHAINS
TOTAL VARIATION CUTOFF IN BIRTH-AND-DEATH CHAINS JIAN DING, EYAL LUBETZKY AND YUVAL PERES ABSTRACT. The cutoff phenomenon describes a case where a Markov chain exhibits a sharp transition in its convergence
More informationCPSC 540: Machine Learning
CPSC 540: Machine Learning Mark Schmidt University of British Columbia Winter 2019 Last Time: Monte Carlo Methods If we want to approximate expectations of random functions, E[g(x)] = g(x)p(x) or E[g(x)]
More informationRepresentation of Markov chains. Stochastic stability AIMS 2014
Markov chain model Joint with J. Jost and M. Kell Max Planck Institute for Mathematics in the Sciences Leipzig 0th July 204 AIMS 204 We consider f : M! M to be C r for r 0 and a small perturbation parameter
More information1 Continuous-time chains, finite state space
Université Paris Diderot 208 Markov chains Exercises 3 Continuous-time chains, finite state space Exercise Consider a continuous-time taking values in {, 2, 3}, with generator 2 2. 2 2 0. Draw the diagramm
More informationAnd now for something completely different
And now for something completely different David Aldous July 27, 2012 I don t do constants. Rick Durrett. I will describe a topic with the paradoxical features Constants matter it s all about the constants.
More informationMean-field dual of cooperative reproduction
The mean-field dual of systems with cooperative reproduction joint with Tibor Mach (Prague) A. Sturm (Göttingen) Friday, July 6th, 2018 Poisson construction of Markov processes Let (X t ) t 0 be a continuous-time
More informationLecture 21. David Aldous. 16 October David Aldous Lecture 21
Lecture 21 David Aldous 16 October 2015 In continuous time 0 t < we specify transition rates or informally P(X (t+δ)=j X (t)=i, past ) q ij = lim δ 0 δ P(X (t + dt) = j X (t) = i) = q ij dt but note these
More informationHomework set 3 - Solutions
Homework set 3 - Solutions Math 495 Renato Feres Problems 1. (Text, Exercise 1.13, page 38.) Consider the Markov chain described in Exercise 1.1: The Smiths receive the paper every morning and place it
More informationARCC WORKSHOP: SHARP THRESHOLDS FOR MIXING TIMES SCRIBES: SHARAD GOEL, ASAF NACHMIAS, PETER RALPH AND DEVAVRAT SHAH
ARCC WORKSHOP: SHARP THRESHOLDS FOR MIXING TIMES SCRIBES: SHARAD GOEL, ASAF NACHMIAS, PETER RALPH AND DEVAVRAT SHAH Day I 1. Speaker I: Persi Diaconis Definition 1 (Sharp Threshold). A Markov chain with
More informationApproximate Inference
Approximate Inference Simulation has a name: sampling Sampling is a hot topic in machine learning, and it s really simple Basic idea: Draw N samples from a sampling distribution S Compute an approximate
More informationMarkov Independence (Continued)
Markov Independence (Continued) As an Undirected Graph Basic idea: Each variable V is represented as a vertex in an undirected graph G = (V(G), E(G)), with set of vertices V(G) and set of edges E(G) the
More information18.175: Lecture 30 Markov chains
18.175: Lecture 30 Markov chains Scott Sheffield MIT Outline Review what you know about finite state Markov chains Finite state ergodicity and stationarity More general setup Outline Review what you know
More informationRandom walk on groups
Random walk on groups Bob Hough September 22, 2015 Bob Hough Random walk on groups September 22, 2015 1 / 12 Set-up G a topological (finite) group, with Borel sigma algebra B P(G) the set of Borel probability
More informationprocess on the hierarchical group
Intertwining of Markov processes and the contact process on the hierarchical group April 27, 2010 Outline Intertwining of Markov processes Outline Intertwining of Markov processes First passage times of
More informationRandom Walks on Finite Quantum Groups
Random Walks on Finite Quantum Groups Diaconis-Shahshahani Upper Bound Lemma for Finite Quantum Groups J.P. McCarthy Cork Institute of Technology 17 May 2017 Seoul National University The Classical Problem
More informationAdiabatic times for Markov chains and applications
Adiabatic times for Markov chains and applications Kyle Bradford and Yevgeniy Kovchegov Department of Mathematics Oregon State University Corvallis, OR 97331-4605, USA bradfork, kovchegy @math.oregonstate.edu
More informationMIXING TIMES OF RANDOM WALKS ON DYNAMIC CONFIGURATION MODELS
MIXING TIMES OF RANDOM WALKS ON DYNAMIC CONFIGURATION MODELS Frank den Hollander Leiden University, The Netherlands New Results and Challenges with RWDRE, 11 13 January 2017, Heilbronn Institute for Mathematical
More informationOn random walks. i=1 U i, where x Z is the starting point of
On random walks Random walk in dimension. Let S n = x+ n i= U i, where x Z is the starting point of the random walk, and the U i s are IID with P(U i = +) = P(U n = ) = /2.. Let N be fixed (goal you want
More informationXVI International Congress on Mathematical Physics
Aug 2009 XVI International Congress on Mathematical Physics Underlying geometry: finite graph G=(V,E ). Set of possible configurations: V (assignments of +/- spins to the vertices). Probability of a configuration
More informationDiaconis Shahshahani Upper Bound Lemma for Finite Quantum Groups
Diaconis Shahshahani Upper Bound Lemma for Finite Quantum Groups J.P. McCarthy Cork Institute of Technology 30 August 2018 Irish Mathematical Society Meeting, UCD Finite Classical Groups aka Finite Groups
More informationU Logo Use Guidelines
Information Theory Lecture 3: Applications to Machine Learning U Logo Use Guidelines Mark Reid logo is a contemporary n of our heritage. presents our name, d and our motto: arn the nature of things. authenticity
More informationConsistency of the maximum likelihood estimator for general hidden Markov models
Consistency of the maximum likelihood estimator for general hidden Markov models Jimmy Olsson Centre for Mathematical Sciences Lund University Nordstat 2012 Umeå, Sweden Collaborators Hidden Markov models
More informationEffective dynamics for the (overdamped) Langevin equation
Effective dynamics for the (overdamped) Langevin equation Frédéric Legoll ENPC and INRIA joint work with T. Lelièvre (ENPC and INRIA) Enumath conference, MS Numerical methods for molecular dynamics EnuMath
More informationPattern Recognition and Machine Learning. Bishop Chapter 11: Sampling Methods
Pattern Recognition and Machine Learning Chapter 11: Sampling Methods Elise Arnaud Jakob Verbeek May 22, 2008 Outline of the chapter 11.1 Basic Sampling Algorithms 11.2 Markov Chain Monte Carlo 11.3 Gibbs
More informationA CLASS OF MARKOV CHAINS WITH NO SPECTRAL GAP
PROCEEDINGS OF THE AMERICAN MATHEMATICAL SOCIETY Volume 4, Number, December 03, Pages 437 436 S 000-9939(03)697-7 Article electronically published on August 6, 03 A CLASS OF MARKOV CHAINS WITH NO SPECTRAL
More informationExchangeability. Peter Orbanz. Columbia University
Exchangeability Peter Orbanz Columbia University PARAMETERS AND PATTERNS Parameters P(X θ) = Probability[data pattern] 3 2 1 0 1 2 3 5 0 5 Inference idea data = underlying pattern + independent noise Peter
More informationLecture 6: Markov Chain Monte Carlo
Lecture 6: Markov Chain Monte Carlo D. Jason Koskinen koskinen@nbi.ku.dk Photo by Howard Jackman University of Copenhagen Advanced Methods in Applied Statistics Feb - Apr 2016 Niels Bohr Institute 2 Outline
More informationErgodic Theorems. Samy Tindel. Purdue University. Probability Theory 2 - MA 539. Taken from Probability: Theory and examples by R.
Ergodic Theorems Samy Tindel Purdue University Probability Theory 2 - MA 539 Taken from Probability: Theory and examples by R. Durrett Samy T. Ergodic theorems Probability Theory 1 / 92 Outline 1 Definitions
More informationZig-Zag Monte Carlo. Delft University of Technology. Joris Bierkens February 7, 2017
Zig-Zag Monte Carlo Delft University of Technology Joris Bierkens February 7, 2017 Joris Bierkens (TU Delft) Zig-Zag Monte Carlo February 7, 2017 1 / 33 Acknowledgements Collaborators Andrew Duncan Paul
More informationAdvanced Sampling Algorithms
+ Advanced Sampling Algorithms + Mobashir Mohammad Hirak Sarkar Parvathy Sudhir Yamilet Serrano Llerena Advanced Sampling Algorithms Aditya Kulkarni Tobias Bertelsen Nirandika Wanigasekara Malay Singh
More informationcurvature, mixing, and entropic interpolation Simons Feb-2016 and CSE 599s Lecture 13
curvature, mixing, and entropic interpolation Simons Feb-2016 and CSE 599s Lecture 13 James R. Lee University of Washington Joint with Ronen Eldan (Weizmann) and Joseph Lehec (Paris-Dauphine) Markov chain
More informationCOUPLING TIMES FOR RANDOM WALKS WITH INTERNAL STATES
COUPLING TIMES FOR RANDOM WALKS WITH INTERNAL STATES ELYSE AZORR, SAMUEL J. GHITELMAN, RALPH MORRISON, AND GREG RICE ADVISOR: YEVGENIY KOVCHEGOV OREGON STATE UNIVERSITY ABSTRACT. Using coupling techniques,
More informationMinicourse on: Markov Chain Monte Carlo: Simulation Techniques in Statistics
Minicourse on: Markov Chain Monte Carlo: Simulation Techniques in Statistics Eric Slud, Statistics Program Lecture 1: Metropolis-Hastings Algorithm, plus background in Simulation and Markov Chains. Lecture
More informationIntroduction to Machine Learning CMU-10701
Introduction to Machine Learning CMU-10701 Markov Chain Monte Carlo Methods Barnabás Póczos Contents Markov Chain Monte Carlo Methods Sampling Rejection Importance Hastings-Metropolis Gibbs Markov Chains
More informationConductance and Rapidly Mixing Markov Chains
Conductance and Rapidly Mixing Markov Chains Jamie King james.king@uwaterloo.ca March 26, 2003 Abstract Conductance is a measure of a Markov chain that quantifies its tendency to circulate around its states.
More informationMaster s Written Examination
Master s Written Examination Option: Statistics and Probability Spring 05 Full points may be obtained for correct answers to eight questions Each numbered question (which may have several parts) is worth
More informationBayesian Nonparametrics
Bayesian Nonparametrics Peter Orbanz Columbia University PARAMETERS AND PATTERNS Parameters P(X θ) = Probability[data pattern] 3 2 1 0 1 2 3 5 0 5 Inference idea data = underlying pattern + independent
More informationLecture 7. µ(x)f(x). When µ is a probability measure, we say µ is a stationary distribution.
Lecture 7 1 Stationary measures of a Markov chain We now study the long time behavior of a Markov Chain: in particular, the existence and uniqueness of stationary measures, and the convergence of the distribution
More informationCPSC 540: Machine Learning
CPSC 540: Machine Learning Mark Schmidt University of British Columbia Winter 2018 Last Time: Monte Carlo Methods If we want to approximate expectations of random functions, E[g(x)] = g(x)p(x) or E[g(x)]
More informationConvex Optimization CMU-10725
Convex Optimization CMU-10725 Simulated Annealing Barnabás Póczos & Ryan Tibshirani Andrey Markov Markov Chains 2 Markov Chains Markov chain: Homogen Markov chain: 3 Markov Chains Assume that the state
More informationLECTURE 3. Last time:
LECTURE 3 Last time: Mutual Information. Convexity and concavity Jensen s inequality Information Inequality Data processing theorem Fano s Inequality Lecture outline Stochastic processes, Entropy rate
More informationLecture 3: September 10
CS294 Markov Chain Monte Carlo: Foundations & Applications Fall 2009 Lecture 3: September 10 Lecturer: Prof. Alistair Sinclair Scribes: Andrew H. Chan, Piyush Srivastava Disclaimer: These notes have not
More informationCoupling. 2/3/2010 and 2/5/2010
Coupling 2/3/2010 and 2/5/2010 1 Introduction Consider the move to middle shuffle where a card from the top is placed uniformly at random at a position in the deck. It is easy to see that this Markov Chain
More informationCUT-OFF FOR QUANTUM RANDOM WALKS. 1. Convergence of quantum random walks
CUT-OFF FOR QUANTUM RANDOM WALKS AMAURY FRESLON Abstract. We give an introduction to the cut-off phenomenon for random walks on classical and quantum compact groups. We give an example involving random
More informationConvergence to equilibrium of Markov processes (eventually piecewise deterministic)
Convergence to equilibrium of Markov processes (eventually piecewise deterministic) A. Guillin Université Blaise Pascal and IUF Rennes joint works with D. Bakry, F. Barthe, F. Bolley, P. Cattiaux, R. Douc,
More informationStatistics & Data Sciences: First Year Prelim Exam May 2018
Statistics & Data Sciences: First Year Prelim Exam May 2018 Instructions: 1. Do not turn this page until instructed to do so. 2. Start each new question on a new sheet of paper. 3. This is a closed book
More informationCONVERGENCE THEOREM FOR FINITE MARKOV CHAINS. Contents
CONVERGENCE THEOREM FOR FINITE MARKOV CHAINS ARI FREEDMAN Abstract. In this expository paper, I will give an overview of the necessary conditions for convergence in Markov chains on finite state spaces.
More informationProject Two. James K. Peterson. March 26, Department of Biological Sciences and Department of Mathematical Sciences Clemson University
Project Two James K. Peterson Department of Biological Sciences and Department of Mathematical Sciences Clemson University March 26, 2019 Outline 1 Cooling Models 2 Estimating the Cooling Rate k 3 Typical
More informationCS 343: Artificial Intelligence
CS 343: Artificial Intelligence Bayes Nets: Sampling Prof. Scott Niekum The University of Texas at Austin [These slides based on those of Dan Klein and Pieter Abbeel for CS188 Intro to AI at UC Berkeley.
More informationSpectra of Large Random Stochastic Matrices & Relaxation in Complex Systems
Spectra of Large Random Stochastic Matrices & Relaxation in Complex Systems Reimer Kühn Disordered Systems Group Department of Mathematics, King s College London Random Graphs and Random Processes, KCL
More informationMATH 320 INHOMOGENEOUS LINEAR SYSTEMS OF DIFFERENTIAL EQUATIONS
MATH 2 INHOMOGENEOUS LINEAR SYSTEMS OF DIFFERENTIAL EQUATIONS W To find a particular solution for a linear inhomogeneous system of differential equations x Ax = ft) or of a mechanical system with external
More informationMarkov Chains, Random Walks on Graphs, and the Laplacian
Markov Chains, Random Walks on Graphs, and the Laplacian CMPSCI 791BB: Advanced ML Sridhar Mahadevan Random Walks! There is significant interest in the problem of random walks! Markov chain analysis! Computer
More information. α β γ δ ε ζ η θ ι κ λ μ Aμ ν(x) ξ ο π ρ ς σ τ υ φ χ ψ ω. Α Β Γ Δ Ε Ζ Η Θ Ι Κ Λ Μ Ν Ξ Ο Π Ρ Σ Τ Υ Φ Χ Ψ Ω. Friday April 1 ± ǁ
. α β γ δ ε ζ η θ ι κ λ μ Aμ ν(x) ξ ο π ρ ς σ τ υ φ χ ψ ω. Α Β Γ Δ Ε Ζ Η Θ Ι Κ Λ Μ Ν Ξ Ο Π Ρ Σ Τ Υ Φ Χ Ψ Ω Friday April 1 ± ǁ 1 Chapter 5. Photons: Covariant Theory 5.1. The classical fields 5.2. Covariant
More informationInformation Retrieval
Introduction to Information Retrieval Lecture 11: Probabilistic Information Retrieval 1 Outline Basic Probability Theory Probability Ranking Principle Extensions 2 Basic Probability Theory For events A
More informationMarkov Chains. X(t) is a Markov Process if, for arbitrary times t 1 < t 2 <... < t k < t k+1. If X(t) is discrete-valued. If X(t) is continuous-valued
Markov Chains X(t) is a Markov Process if, for arbitrary times t 1 < t 2
More informationCONVERGENCE RATE OF MCMC AND SIMULATED ANNEAL- ING WITH APPLICATION TO CLIENT-SERVER ASSIGNMENT PROBLEM
Stochastic Analysis and Applications CONVERGENCE RATE OF MCMC AND SIMULATED ANNEAL- ING WITH APPLICATION TO CLIENT-SERVER ASSIGNMENT PROBLEM THUAN DUONG-BA, THINH NGUYEN, BELLA BOSE School of Electrical
More informationThe Theory behind PageRank
The Theory behind PageRank Mauro Sozio Telecom ParisTech May 21, 2014 Mauro Sozio (LTCI TPT) The Theory behind PageRank May 21, 2014 1 / 19 A Crash Course on Discrete Probability Events and Probability
More informationThe Markov Chain Monte Carlo Method
The Markov Chain Monte Carlo Method Idea: define an ergodic Markov chain whose stationary distribution is the desired probability distribution. Let X 0, X 1, X 2,..., X n be the run of the chain. The Markov
More informationCS 188: Artificial Intelligence Spring Announcements
CS 188: Artificial Intelligence Spring 2011 Lecture 18: HMMs and Particle Filtering 4/4/2011 Pieter Abbeel --- UC Berkeley Many slides over this course adapted from Dan Klein, Stuart Russell, Andrew Moore
More informationG(t) := i. G(t) = 1 + e λut (1) u=2
Note: a conjectured compactification of some finite reversible MCs There are two established theories which concern different aspects of the behavior of finite state Markov chains as the size of the state
More informationPCMI LECTURE NOTES ON PROPERTY (T ), EXPANDER GRAPHS AND APPROXIMATE GROUPS (PRELIMINARY VERSION)
PCMI LECTURE NOTES ON PROPERTY (T ), EXPANDER GRAPHS AND APPROXIMATE GROUPS (PRELIMINARY VERSION) EMMANUEL BREUILLARD 1. Lecture 1, Spectral gaps for infinite groups and non-amenability The final aim of
More informationMCMC and Gibbs Sampling. Sargur Srihari
MCMC and Gibbs Sampling Sargur srihari@cedar.buffalo.edu 1 Topics 1. Markov Chain Monte Carlo 2. Markov Chains 3. Gibbs Sampling 4. Basic Metropolis Algorithm 5. Metropolis-Hastings Algorithm 6. Slice
More informationA GIBBS SAMPLER ON THE N-SIMPLEX. By Aaron Smith Stanford University
Submitted to the Annals of Applied Probability arxiv: math.pr/0290964 A GIBBS SAMPLER ON THE N-SIMPLEX By Aaron Smith Stanford University We determine the mixing time of a simple Gibbs sampler on the unit
More informationStochastic Optimization Algorithms Beyond SG
Stochastic Optimization Algorithms Beyond SG Frank E. Curtis 1, Lehigh University involving joint work with Léon Bottou, Facebook AI Research Jorge Nocedal, Northwestern University Optimization Methods
More informationMA1021 Calculus I B Term, Sign:
MA1021 Calculus I B Term, 2014 Final Exam Print Name: Sign: Write up your solutions neatly and show all your work. 1. (28 pts) Compute each of the following derivatives: You do not have to simplify your
More informationCSCI 1951-G Optimization Methods in Finance Part 12: Variants of Gradient Descent
CSCI 1951-G Optimization Methods in Finance Part 12: Variants of Gradient Descent April 27, 2018 1 / 32 Outline 1) Moment and Nesterov s accelerated gradient descent 2) AdaGrad and RMSProp 4) Adam 5) Stochastic
More informationsystem CWI, Amsterdam May 21, 2008 Dynamic Analysis Seminar Vrije Universiteit
CWI, Amsterdam heijster@cwi.nl May 21, 2008 Dynamic Analysis Seminar Vrije Universiteit Joint work: A. Doelman (CWI/UvA), T.J. Kaper (BU), K. Promislow (MSU) Outline 1 2 3 4 Outline 1 2 3 4 Paradigm U
More informationPolynomial-time counting and sampling of two-rowed contingency tables
Polynomial-time counting and sampling of two-rowed contingency tables Martin Dyer and Catherine Greenhill School of Computer Studies University of Leeds Leeds LS2 9JT UNITED KINGDOM Abstract In this paper
More informationExercises with solutions (Set D)
Exercises with solutions Set D. A fair die is rolled at the same time as a fair coin is tossed. Let A be the number on the upper surface of the die and let B describe the outcome of the coin toss, where
More informationCut-off and Escape Behaviors for Birth and Death Chains on Trees
ALEA, Lat. Am. J. Probab. Math. Stat. 8, 49 6 0 Cut-off and Escape Behaviors for Birth and Death Chains on Trees Olivier Bertoncini Department of Physics, University of Aveiro, 380-93 Aveiro, Portugal.
More informationAssignment 5 Bounding Complexities KEY
Assignment 5 Bounding Complexities KEY Print this sheet and fill in your answers. Please staple the sheets together. Turn in at the beginning of class on Friday, September 16. There are links in the examples
More informationLecture 7 and 8: Markov Chain Monte Carlo
Lecture 7 and 8: Markov Chain Monte Carlo 4F13: Machine Learning Zoubin Ghahramani and Carl Edward Rasmussen Department of Engineering University of Cambridge http://mlg.eng.cam.ac.uk/teaching/4f13/ Ghahramani
More informationBasic math for biology
Basic math for biology Lei Li Florida State University, Feb 6, 2002 The EM algorithm: setup Parametric models: {P θ }. Data: full data (Y, X); partial data Y. Missing data: X. Likelihood and maximum likelihood
More informationBayes Net Representation. CS 188: Artificial Intelligence. Approximate Inference: Sampling. Variable Elimination. Sampling.
188: Artificial Intelligence Bayes Nets: ampling Bayes Net epresentation A directed, acyclic graph, one node per random variable A conditional probability table (PT) for each node A collection of distributions
More informationMATH 564/STAT 555 Applied Stochastic Processes Homework 2, September 18, 2015 Due September 30, 2015
ID NAME SCORE MATH 56/STAT 555 Applied Stochastic Processes Homework 2, September 8, 205 Due September 30, 205 The generating function of a sequence a n n 0 is defined as As : a ns n for all s 0 for which
More information