comparison of random walk based techniques for estimating
play

Comparison of Random Walk based techniques for estimating network - PowerPoint PPT Presentation

Jan 14, 2023 •152 likes •1.18k views

Comparison of Random Walk based techniques for estimating network averages Jithin K. Sreedharan INRIA, France Arun Kadavankandy Konstantin Avrachenkov Vivek S. Borkar INRIA, France INRIA, France IIT Bombay, India CSoNet 2016, August 2

  1. Reinforcement Learning technique (contd.) Stochastic Approximation Seed set For each node 𝑗 in Algorithm Jithin K. Sreedharan (jithin.sreedharan@inria.fr) 9

  2. Reinforcement Learning technique (contd.) Stochastic Approximation Seed set For each node 𝑗 in Function sum Algorithm inside a tour Cost function Jithin K. Sreedharan (jithin.sreedharan@inria.fr) 9

  3. Reinforcement Learning technique (contd.) Stochastic Approximation Seed set For each node 𝑗 in Function sum Algorithm inside a tour Cost function ……. sample 1 sample 2 sample 𝑙 Jithin K. Sreedharan (jithin.sreedharan@inria.fr) 9

  4. Reinforcement Learning technique (contd.) Stochastic Approximation Seed set For each node 𝑗 in Function sum Algorithm inside a tour Cost function ……. sample 1 sample 2 sample 𝑙 Jithin K. Sreedharan (jithin.sreedharan@inria.fr) 9

  5. Reinforcement Learning technique (contd.) Stochastic Approximation Seed set For each node 𝑗 in Function sum Algorithm inside a tour Cost function Jithin K. Sreedharan (jithin.sreedharan@inria.fr) 9

  6. Which Random Walk method to select ? Jithin K. Sreedharan (jithin.sreedharan@inria.fr) 10

  7. Which Random Walk method to select ?  Mixing time Not a good criterion here due to burn-in period. Jithin K. Sreedharan (jithin.sreedharan@inria.fr) 10

  8. Which Random Walk method to select ?  Mixing time Not a good criterion here due to burn-in period. X k : rejected sample : accepted sample ……. X 1 X 2 Burn-in period Approximate stationary regime Jithin K. Sreedharan (jithin.sreedharan@inria.fr) 10

  9. Which Random Walk method to select ?  Mixing time Not a good criterion here due to burn-in period. X k : rejected sample : accepted sample ……. X 1 X 2 Burn-in period Approximate stationary regime Reinforcement Learning technique does not require burn-in period Jithin K. Sreedharan (jithin.sreedharan@inria.fr) 10

  10. Which Random Walk method to select ?  Mixing time Not a good criterion here due to burn-in period.  Efficiency of the estimator: How many samples are needed to achieve certain accuracy Jithin K. Sreedharan (jithin.sreedharan@inria.fr) 10

  11. Asymptotic Variance Jithin K. Sreedharan (jithin.sreedharan@inria.fr) 11

  12. Asymptotic Variance Asymptotic variance of the estimator Jithin K. Sreedharan (jithin.sreedharan@inria.fr) 11

  13. Asymptotic Variance Asymptotic variance of the estimator Also from Central Limit Theorem equivalent Jithin K. Sreedharan (jithin.sreedharan@inria.fr) 11

  14. Asymptotic Variance (contd.) Jithin K. Sreedharan (jithin.sreedharan@inria.fr) 12

  15. Asymptotic Variance (contd.)  For Metropolis-Hastings Sampling, where Fundamental matrix of Markov chain Jithin K. Sreedharan (jithin.sreedharan@inria.fr) 12

  16. Asymptotic Variance (contd.)  For Metropolis-Hastings Sampling, where Fundamental matrix of Markov chain  For Respondent Driven Sampling, Jithin K. Sreedharan (jithin.sreedharan@inria.fr) 12

  17. Asymptotic Variance (contd.)  For Metropolis-Hastings Sampling, where Fundamental matrix of Markov chain  For Respondent Driven Sampling,  For Reinforcement Learning based sampling, Jithin K. Sreedharan (jithin.sreedharan@inria.fr) 12

  18. Numerical Studies Jithin K. Sreedharan (jithin.sreedharan@inria.fr) 13

  19. Numerical Studies Normalized Root Mean Square Error (NRMSE) vs Budget B Jithin K. Sreedharan (jithin.sreedharan@inria.fr) 13

  20. Numerical Studies Normalized Root Mean Square Error (NRMSE) vs Budget B Jithin K. Sreedharan (jithin.sreedharan@inria.fr) 13

  21. Numerical Studies Normalized Root Mean Square Error (NRMSE) vs Budget B Why MSE ? Jithin K. Sreedharan (jithin.sreedharan@inria.fr) 13

  22. Numerical Studies Normalized Root Mean Square Error (NRMSE) vs Budget B Why MSE ? Jithin K. Sreedharan (jithin.sreedharan@inria.fr) 13

  23. Numerical Studies Normalized Root Mean Square Error (NRMSE) vs Budget B Why MSE ? Budget B: number of allowed samples Jithin K. Sreedharan (jithin.sreedharan@inria.fr) 13

  24. Les Misérables network Jithin K. Sreedharan (jithin.sreedharan@inria.fr) 14

  25. Les Misérables network Number of nodes: 77, number of edges: 254. Jithin K. Sreedharan (jithin.sreedharan@inria.fr) 14

  26. Les Misérables network Number of nodes: 77, number of edges: 254. Jithin K. Sreedharan (jithin.sreedharan@inria.fr) 14

  27. Les Misérables network Number of nodes: 77, number of edges: 254. Jithin K. Sreedharan (jithin.sreedharan@inria.fr) 14

  28. Les Misérables network contd. Jithin K. Sreedharan (jithin.sreedharan@inria.fr) 15

  29. Les Misérables network contd. Jithin K. Sreedharan (jithin.sreedharan@inria.fr) 15

  30. Les Misérables network contd. Jithin K. Sreedharan (jithin.sreedharan@inria.fr) 15

  31. Les Misérables network contd. Jithin K. Sreedharan (jithin.sreedharan@inria.fr) 16

  32. Les Misérables network contd. Study of asymptotic variance Jithin K. Sreedharan (jithin.sreedharan@inria.fr) 16

  33. Les Misérables network contd. Study of asymptotic variance Jithin K. Sreedharan (jithin.sreedharan@inria.fr) 16

  34. Les Misérables network contd. Study of asymptotic variance Jithin K. Sreedharan (jithin.sreedharan@inria.fr) 16

  35. Les Misérables network contd. Study of asymptotic variance Jithin K. Sreedharan (jithin.sreedharan@inria.fr) 16

  36. Friendster network Jithin K. Sreedharan (jithin.sreedharan@inria.fr) 17

  37. Friendster network Number of nodes ~ 65K number of edges ~ 1.25M Jithin K. Sreedharan (jithin.sreedharan@inria.fr) 17

  38. Friendster network Number of nodes ~ 65K number of edges ~ 1.25M Jithin K. Sreedharan (jithin.sreedharan@inria.fr) 17

  39. Friendster network Number of nodes ~ 65K number of edges ~ 1.25M Jithin K. Sreedharan (jithin.sreedharan@inria.fr) 17

  40. Friendster network contd. Jithin K. Sreedharan (jithin.sreedharan@inria.fr) 18

  41. Friendster network contd. Stability of sample paths: Jithin K. Sreedharan (jithin.sreedharan@inria.fr) 18

  42. Friendster network contd. Stability of sample paths: single path example Jithin K. Sreedharan (jithin.sreedharan@inria.fr) 18

  43. Friendster network contd. Stability of sample paths: single path example Varying super-node size Jithin K. Sreedharan (jithin.sreedharan@inria.fr) 18

  44. Friendster network contd. Stability of sample paths: single path example Varying super-node size Varying step size Jithin K. Sreedharan (jithin.sreedharan@inria.fr) 18

  45. Conclusions Jithin K. Sreedharan (jithin.sreedharan@inria.fr) 19

  46. Conclusions  Rand Walk based estimators of Jithin K. Sreedharan (jithin.sreedharan@inria.fr) 19

  47. Conclusions  Rand Walk based estimators of  Numerical and theoretical study of Mean Square Error & Asymptotic Variance of  Metropolis-Hastings sampling  Respondent Driven sampling (RDS)  New Reinforcement Learning based sampling (RL) Jithin K. Sreedharan (jithin.sreedharan@inria.fr) 19

  48. Conclusions  Rand Walk based estimators of  Numerical and theoretical study of Mean Square Error & Asymptotic Variance of  Metropolis-Hastings sampling  Respondent Driven sampling (RDS)  New Reinforcement Learning based sampling (RL)  Reinforcement Learning technique:  Tackles disconnected graph  A cross between deterministic iteration and MCMC  Can control the stability of the algorithm with step sizes Jithin K. Sreedharan (jithin.sreedharan@inria.fr) 19

  49. Conclusions  Rand Walk based estimators of  Numerical and theoretical study of Mean Square Error & Asymptotic Variance of  Metropolis-Hastings sampling  Respondent Driven sampling (RDS)  New Reinforcement Learning based sampling (RL)  Reinforcement Learning technique:  Tackles disconnected graph  A cross between deterministic iteration and MCMC  Can control the stability of the algorithm with step sizes  RDS works better. RL technique comparable, yet more stable and no burn-in ! Jithin K. Sreedharan (jithin.sreedharan@inria.fr) 19

Download Presentation
Download Policy: The content available on the website is offered to you 'AS IS' for your personal information and use only. It cannot be commercialized, licensed, or distributed on other websites without prior consent from the author. To download a presentation, simply click this link. If you encounter any difficulties during the download process, it's possible that the publisher has removed the file from their server.

Recommend

Estimating Properties of Social Networks via Random Walk considering Private Nodes Kazuki

Estimating Properties of Social Networks via Random Walk considering Private Nodes Kazuki

KDD 2020 Research track Estimating Properties of Social Networks via Random Walk considering Private Nodes Kazuki Nakajima Kazuyuki Shudo Tokyo Institute of Technology Graph Sampling on Social Networks Purpose: Understand global social

443 views • 21 slides
Short Walks in Higher Dimensions Ghislain McKay Febuary 3, 2015 What is a Random Walk? A path

Short Walks in Higher Dimensions Ghislain McKay Febuary 3, 2015 What is a Random Walk? A path

Short Walks in Higher Dimensions Ghislain McKay Febuary 3, 2015 What is a Random Walk? A path formed by a succession of n steps (of unit length) in random directions. Figure: A 26-step random walk in the plane What is a Random Walk? A path

474 views • 44 slides
Random Walks Will Perkins February 5, 2013 Simple Random Walk S 0 = 0, S n = X 1 + X 2 + . . . X

Random Walks Will Perkins February 5, 2013 Simple Random Walk S 0 = 0, S n = X 1 + X 2 + . . . X

Random Walks Will Perkins February 5, 2013 Simple Random Walk S 0 = 0, S n = X 1 + X 2 + . . . X n , where X i s are iid 1 with probability p and 1 p Combinatorics Calculating the number of simple random walk paths from (0 , 0) to ( n

241 views • 12 slides
and Size of Social Networks via Random Walk Stephen J. Hardiman* Liran Katzir Capital Fund

and Size of Social Networks via Random Walk Stephen J. Hardiman* Liran Katzir Capital Fund

Estimating Clustering Coefficients and Size of Social Networks via Random Walk Stephen J. Hardiman* Liran Katzir Capital Fund Management Advanced Technology Labs France Microsoft Research, Israel *Research was conducted while the author was

467 views • 34 slides
Utilizing Macromodels in Floating Random Walk Based Capacitance Extraction Wenjian Yu Department

Utilizing Macromodels in Floating Random Walk Based Capacitance Extraction Wenjian Yu Department

Utilizing Macromodels in Floating Random Walk Based Capacitance Extraction Wenjian Yu Department of Computer Science and Technology, Tsinghua University , Beijing 100084, China yu-wj@tsinghua.edu.cn Based on the paper by W. Yu, B. Zhang, C.

406 views • 24 slides
Prosody-Based Unsupervised Speech Summarization with Two-Layer Mutually Reinforced Random Walk

Prosody-Based Unsupervised Speech Summarization with Two-Layer Mutually Reinforced Random Walk

Prosody-Based Unsupervised Speech Summarization with Two-Layer Mutually Reinforced Random Walk Sujay Kumar Jauhar {sjauhar, yvchen, fmetze}@cs.cmu.edu Yun-Nung (Vivian) Chen The 6th International Joint Conference on Natural Florian Metze

768 views • 38 slides
A random walk based clustering with local re-computations for mobile ad hoc networks Kudireti

A random walk based clustering with local re-computations for mobile ad hoc networks Kudireti

Introduction Algorithm Example execution of the algorithm Properties Simulation results Conclusion and perspectives A random walk based clustering with local re-computations for mobile ad hoc networks Kudireti ABDURUSUL Alain BUI

406 views • 40 slides
Balanced Allocation with Random Walk Based Sampling Dengwang Tang Electrical and Computer

Balanced Allocation with Random Walk Based Sampling Dengwang Tang Electrical and Computer

Introduction Related Work Main Results Discussion References Balanced Allocation with Random Walk Based Sampling Dengwang Tang Electrical and Computer Engineering Department University of Michigan, Ann Arbor Joint work with Vijay

698 views • 46 slides
Estimating Risk under Estimating statistics . . . Linearized techniques Interval Uncertainty:

Estimating Risk under Estimating statistics . . . Linearized techniques Interval Uncertainty:

Computing statistics is . . . Additional problem: . . . Traditional approach . . . Estimating Risk under Estimating statistics . . . Linearized techniques Interval Uncertainty: The exact estimation . . . Need for parallelization Sequential

238 views • 21 slides
RANDOM WALK IN DYNAMIC RANDOM ENVIRONMENT Frank den Hollander Leiden University The Netherlands

RANDOM WALK IN DYNAMIC RANDOM ENVIRONMENT Frank den Hollander Leiden University The Netherlands

RANDOM WALK IN DYNAMIC RANDOM ENVIRONMENT Frank den Hollander Leiden University The Netherlands Joint work with: M. Hil ario, R. dos Santos (Belo Horizonte) V. Sidoravicius, A. Teixeira (Rio de Janeiro) School and Workshop on Random

521 views • 18 slides
Mixing time for a random walk on a ring Stephen Connor Joint work with Michael Bate Aspects of

Mixing time for a random walk on a ring Stephen Connor Joint work with Michael Bate Aspects of

RW on a group Random walk on a ring Problems Mixing time for a random walk on a ring Stephen Connor Joint work with Michael Bate Aspects of Random Walks Durham, April 2014 RW on a group Random walk on a ring Problems Random walks on

512 views • 38 slides
Critical density for Activated Random Walk Lorenzo Taggi Max Planck Institute for Mathematics in

Critical density for Activated Random Walk Lorenzo Taggi Max Planck Institute for Mathematics in

Critical density for Activated Random Walk Critical density for Activated Random Walk Lorenzo Taggi Max Planck Institute for Mathematics in the Sciences Leipzig, Germany June 24, 2014 1 Critical density for Activated Random Walk Outline 1

612 views • 29 slides
Random Walk Based Algorithms for Complex Network Analysis Konstantin Avrachenkov Inria Sophia

Random Walk Based Algorithms for Complex Network Analysis Konstantin Avrachenkov Inria Sophia

Random Walk Based Algorithms for Complex Network Analysis Konstantin Avrachenkov Inria Sophia Antipolis Rescom 2014, 12-16 May, Furiani, Corse Complex networks Main features of complex networks: Sparse topology; Heavy-tail degree

870 views • 48 slides
Statistically Based Model Comparison Techniques H. T. Banks Center for Research in Scientific

Statistically Based Model Comparison Techniques H. T. Banks Center for Research in Scientific

Statistically Based Model Comparison Techniques H. T. Banks Center for Research in Scientific Computation (CRSC) Center for Quantitative Sciences in Biomedicine (CQSB) North Carolina State University Raleigh, NC 27695 C enter for Q uantitative

685 views • 30 slides
A rate of convergence for loop-erased random walk to SLE(2) Michael J. Kozdron University of

A rate of convergence for loop-erased random walk to SLE(2) Michael J. Kozdron University of

A rate of convergence for loop-erased random walk to SLE(2) Michael J. Kozdron University of Regina http://stat.math.uregina.ca/ kozdron/ Conformal maps from probability to physics Centro Stefano Franscini May 24, 2010 Based on joint work

485 views • 24 slides
Random Walks Conditioned to Stay Positive Bob Keener Let S n be a random walk formed by summing

Random Walks Conditioned to Stay Positive Bob Keener Let S n be a random walk formed by summing

Random Walks Conditioned to Stay Positive Bob Keener Let S n be a random walk formed by summing i.i.d. integer valued random variables X i , i 1 : S n = X 1 + + X n . If the drift EX i is negative, then S n as n . If

145 views • 13 slides
Markov Chains and MCMC CompSci 590.04 Instructor:

Markov Chains and MCMC CompSci 590.04 Instructor:

Markov Chains and MCMC CompSci 590.04 Instructor: AshwinMachanavajjhala Lecture 5 : 590.04 Fall 15 1 Recap: Monte Carlo Method If U is a

715 views • 12 slides
Textures I Week 9, Wed Mar 14 http://www.ugrad.cs.ubc.ca/~cs314/Vjan2007 Reading for Today and

Textures I Week 9, Wed Mar 14 http://www.ugrad.cs.ubc.ca/~cs314/Vjan2007 Reading for Today and

University of British Columbia CPSC 314 Computer Graphics Jan-Apr 2007 Tamara Munzner Textures I Week 9, Wed Mar 14 http://www.ugrad.cs.ubc.ca/~cs314/Vjan2007 Reading for Today and Next Time FCG Chap 11 Texture Mapping except 11.8

616 views • 58 slides
Evidence for anisotropy of cosmic acceleration Subir Sarkar Colin, Mohayaee, Rameez & S.S.,

Evidence for anisotropy of cosmic acceleration Subir Sarkar Colin, Mohayaee, Rameez & S.S.,

Evidence for anisotropy of cosmic acceleration Subir Sarkar Colin, Mohayaee, Rameez & S.S., A&A 631: L13, 2019 Rudolf Peierls Centre for Theoretical Physics 2019 1929 2016 2001 Seminar , U , Universi sity o y of G Geneva, 1 , 17

443 views • 42 slides
Search for a Dark Photon: proposal for the experiment at VEPP3. I.Rachek, B.Wojtsekhowski,

Search for a Dark Photon: proposal for the experiment at VEPP3. I.Rachek, B.Wojtsekhowski,

Search for a Dark Photon: proposal for the experiment at VEPP3. I.Rachek, B.Wojtsekhowski, D.Nikolenko IEBWorkshop Cornell University June 18, 2015 I.Rachek dark photon at VEPP-3 June 18, 2015 1 Latest results of dark photon searches

455 views • 26 slides
How to Encipher Messages on a Small Domain Deterministic Encryption and the Thorp Shuffle Ben

How to Encipher Messages on a Small Domain Deterministic Encryption and the Thorp Shuffle Ben

How to Encipher Messages on a Small Domain Deterministic Encryption and the Thorp Shuffle Ben Morris Phil Rogaway Till Stegers University of California, Davis University of California, Davis Dept of Mathematics Dept of Computer

1.01k views • 25 slides
Markov chain Monte Carlo Reminder Need to sample large, non-standard distributions: Markov

Markov chain Monte Carlo Reminder Need to sample large, non-standard distributions: Markov

Markov chain Monte Carlo Reminder Need to sample large, non-standard distributions: Markov chain Monte Carlo (MCMC) Gibbs and MetropolisHastings S P ( |D ) = P ( D| ) P ( ) P ( x |D ) 1 Slice sampling P

465 views • 8 slides
Stochastic Simulation Discrete simulation/event-by-event Bo Friis Nielsen Institute of

Stochastic Simulation Discrete simulation/event-by-event Bo Friis Nielsen Institute of

Stochastic Simulation Discrete simulation/event-by-event Bo Friis Nielsen Institute of Mathematical Modelling Technical University of Denmark 2800 Kgs. Lyngby Denmark Email: bfni@dtu.dk Discrete event simulation Discrete event

495 views • 25 slides
The Stochastic EM Algorithm for Censored Mixed Models Ian Marschner Macquarie University, Sydney

The Stochastic EM Algorithm for Censored Mixed Models Ian Marschner Macquarie University, Sydney

The Stochastic EM Algorithm for Censored Mixed Models Ian Marschner Macquarie University, Sydney 10th International Conference on MCQMC Methods in Scientific Computing, Sydney, 2012 1 / 18 INTRODUCTION Stochastic EM algorithm is a Monte

441 views • 18 slides

More recommend