mcmc based machine learning
play

MCMC based machine learning a . (Bayesian Model Averaging) Nicos - PowerPoint PPT Presentation

Dec 07, 2023 •171 likes •393 views

MCMC based machine learning a . (Bayesian Model Averaging) Nicos Angelopoulos n.angelopoulos@ed.ac.uk School of Biological Sciences Biochemistry Group University of Edinburgh, Scotland, UK. a Collaborative work with James Cussens, York

  1. MCMC based machine learning a . (Bayesian Model Averaging) Nicos Angelopoulos n.angelopoulos@ed.ac.uk School of Biological Sciences Biochemistry Group University of Edinburgh, Scotland, UK. a Collaborative work with James Cussens, York University, jc@cs.york.ac.uk P .E.L. 2006 – p.1

  2. MCMC Overview Class of sampling algorithms that estimate a posterior distribution. Markov chain construct a chain of visited values, M 1 , M 2 , . . . , M n , by proposing M ∗ from M i , with probability q ( M ∗ , M i ) . Use prior knowledge, p ( M ∗ ) and relative likelihood of the two values, p ( D | M ∗ ) /p ( D | M i ) to decide chain construction. Monte Carlo Use the chain to approximate the posterior p ( M | D ) . P .E.L. 2006 – p.2

  3. Bayesian learning with MCMC Given some data D and a class of statistical models M ( M ∈ M ) that can express relations in the data, use MCMC to approximate normalisation factor in Bayes’ theorem p ( D | M ) p ( M ) p ( M | D ) = � M p ( D | M ) p ( M ) p ( M ) is the prior probability of each model p ( D | M ) the likelihood (how well the model fits the data) p ( M | D ) the posterior P .E.L. 2006 – p.3

  4. Example: Data smoker bronchitis l_cancer person 1 y y n person 2 y n n person 3 y y y person 4 n y n person 5 n n n P .E.L. 2006 – p.4

  5. Example: Models S B 1 B L [b-[],l-[],s-[]] S B 2 B L [b-[s],l-[],s-[]] . . . S B 24 [b-[s],l-[b,s],s-[]] B L P .E.L. 2006 – p.5

  6. Example: Objective P(Bx) . . . B1 B2 B3 B4 . . . B24 � p ( B x ) = 1 B x P .E.L. 2006 – p.6

  7. Metropolis-Hastings (M-H) MCMC 0. Set i = 0 and find M 0 using the prior. 1. From M i produce a candidate model M ∗ . Let the probability of reaching M ∗ be q ( M ∗ , M i ) . 2. Let � q ( M ∗ , M i ) P ( D | M ∗ ) P ( M ∗ ) � α ( M i , M ∗ ) = min q ( M i , M ∗ ) P ( D | M i ) P ( M i ) , 1 � M ∗ with probability α ( M i , M ∗ ) M i +1 = M i with probability 1 − α ( M i , M ∗ ) 3. If i reached limit then terminate, else set i = i + 1 and repeat from 1. P .E.L. 2006 – p.7

  8. Example: MCMC Markov Chain: M 1 B 3 P .E.L. 2006 – p.8

  9. Example: MCMC Markov Chain: M 1 , M 2 B 3 , B 3 P .E.L. 2006 – p.8

  10. Example: MCMC Markov Chain: M 1 , M 2 , M 3 , M 4 , M 5 , . . . B 3 , B 3 , B 10 , B 3 , B 24 , . . . P .E.L. 2006 – p.8

  11. Example: MCMC Markov Chain: M 1 , M 2 , M 3 , M 4 , M 5 , . . . B 3 , B 3 , B 10 , B 3 , B 24 , . . . Monte Carlo: #( B k ) p ( B k ) = � B x #( B x ) P .E.L. 2006 – p.8

  12. SLP defined model space ?− bn( [1,2,3], Bn ). G0 Gi Mi M* From M i identify G i then sample forward to M ⋆ . q ( M i , M ⋆ ) is the probability of proposing M ⋆ when M i is the current model. P .E.L. 2006 – p.9

  13. BN Prior bn( OrdNodes, Bn ) :- bn( Nodes, [], Bn ). bn( [], _PotPar, [] ). bn( [H|T], PotPar, [H-SelParOfH|RemBn] ) :- select_parents( PotPar, H, SelParOfH ), bn( T, [H|PotPar], RemBn ). select_parents( [], [] ). select_parents( [H|T], Pa ) :- include_element( H, Pa, RemPa ), select_parents( T, TPa ). 1/2 : include_element( H, [H|TPa], TPa ). 1/2 : include_element( _H, TPa, TPa ). P .E.L. 2006 – p.10

  14. example BN (Asia) For example ? - bn( [1,2,3,4,5,6,7,8], M ). M = [1-[],2-[1],3-[2,5],4-[],5-[4],6-[4],7-[3],8-[3,6]]. P .E.L. 2006 – p.11

  15. visits and stays P .E.L. 2006 – p.12

  16. Edges recovery With topological ordering constraint and a maximum of 2 parents per node, the algorithm recovers most of the BN arcs in 0.5 M iterations. For example for a .99 cut -off we have : Missing : 2 → 3 (.84) 3 → 7 (.47) Superfluous : 5 → 7 P .E.L. 2006 – p.13

  17. CART priors ? - cart( M ). x2 P split ( η ) = α (1 + d η ) − β =< 1 1 < x1 =< 0 0 < M = node( b, 1, node(a,0,leaf,leaf), leaf ) 1 - Sp: [Sp]: cart( Data, D, A/B, leaf(Data) ). Sp: [Sp]: cart( Data, D, A/B, node(F,V,L,R) ) :- branch( Data, F, V, LData, RData ), D1 is D + 1, NxtSp is A * ((1 + D1) ˆ -B), [NxtSp] : cart( LData, D1, A/B, L ), [NxtSp] : cart( RData, D1, A/B, R ). P .E.L. 2006 – p.14

  18. Experiment Pima Indians Diabetes Database 768 complete entries of 8 variables. Denison et.al. run 250,000 iterations of local perturbations. Their best likelihood model: -343.056 Our experiment run for 250,000 iterations with branch replacing. Parameters: uniform choice proposal, α = . 95 β = . 8 Our best likelihood model: -347.651 P .E.L. 2006 – p.15

  19. Likelihoods trace -340 ’tr_uc_rm_pima_idsd_a0_95b0_8_i250K__s776.llhoods’ -350 -360 -370 -380 -390 -400 -410 -420 0 50000 100000 150000 200000 250000 β = . 8 , α = . 95 , proposal = uniform choice P .E.L. 2006 – p.16

  20. -347.61529077520584 Best likelihood 1:6 best_llhood:vst(37):msclf(145) =< 29.3 >29.3 2:8 11:2 =< 27 >27 =< 166 >166 3:141/5 4:6 12:8 31:7/62 =< 26.3 >26.3 =< 29 >29 5:2 8:8 13:2 28:2 =< 152 >152 =< 54 >54 =< 127 >127 =< 106 >106 6:54/3 7:2/9 9:22/24 10:10/1 14:6 17:3 29:52/19 30:53/92 =< 45.4 >45.4 =< 61 >61 15:129/19 16:1/4 18:1/11 19:8 =< 27 >27 20:5 27:7/5 =< 200 >200 21:7 26:12/1 =< 0.314 >0.314 22:4/1 23:5 =< 32 >32 24:4/6 25:1/6 P .E.L. 2006 – p.17

  21. in Kyoto Models: HMRFs for clustering. Likelihood: design and implement a likelihood -ratio function for HMRFs. Proposal: implement function(s) for reaching proposal model. Application: to real data. SLPs: for more complex priors. P .E.L. 2006 – p.18

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

Advanced Machine Learning MCMC Methods Amit Sethi Electrical Engineering, IIT Bombay Objectives

Advanced Machine Learning MCMC Methods Amit Sethi Electrical Engineering, IIT Bombay Objectives

Advanced Machine Learning MCMC Methods Amit Sethi Electrical Engineering, IIT Bombay Objectives We have talked about: Exact inference in Factor Graphs using Sum- Product algorithm (aka Belief Propagation) Limitations and some

422 views • 24 slides
Markov Chain Monte Carlo (MCMC) Variational methods Milos Hauskrecht milos@cs.pitt.edu

Markov Chain Monte Carlo (MCMC) Variational methods Milos Hauskrecht milos@cs.pitt.edu

CS 3750 Machine Learning Lecture 6 Approximate probabilistic inference: Markov Chain Monte Carlo (MCMC) Variational methods Milos Hauskrecht milos@cs.pitt.edu 5329 Sennott Square CS 3750 Advanced Machine Learning Markov chain

436 views • 21 slides
10b Machine Learning: Symbol-based 10.0 Introduction 10.5 Knowledge and Learning 10.1 A

10b Machine Learning: Symbol-based 10.0 Introduction 10.5 Knowledge and Learning 10.1 A

10b Machine Learning: Symbol-based 10.0 Introduction 10.5 Knowledge and Learning 10.1 A Framework for 10.6 Unsupervised Learning Symbol-based Learning 10.7 Reinforcement Learning 10.2 Version Space Search 10.8 Epilogue and 10.3 The

866 views • 50 slides
10c Machine Learning: Symbol-based 10.0 Introduction 10.5 Knowledge and Learning 10.1 A

10c Machine Learning: Symbol-based 10.0 Introduction 10.5 Knowledge and Learning 10.1 A

10c Machine Learning: Symbol-based 10.0 Introduction 10.5 Knowledge and Learning 10.1 A Framework for 10.6 Unsupervised Learning Symbol-based Learning 10.7 Reinforcement Learning 10.2 Version Space Search 10.8 Epilogue and 10.3 The

605 views • 35 slides
10d Machine Learning: Symbol-based 10.0 Introduction 10.5 Knowledge and Learning 10.1 A

10d Machine Learning: Symbol-based 10.0 Introduction 10.5 Knowledge and Learning 10.1 A

10d Machine Learning: Symbol-based 10.0 Introduction 10.5 Knowledge and Learning 10.1 A Framework for 10.6 Unsupervised Learning Symbol-based Learning 10.7 Reinforcement Learning 10.2 Version Space Search 10.8 Epilogue and 10.3 The

1.05k views • 29 slides
10a Machine Learning: Symbol-based 10.0 Introduction 10.5 Knowledge and Learning 10.1 A

10a Machine Learning: Symbol-based 10.0 Introduction 10.5 Knowledge and Learning 10.1 A

10a Machine Learning: Symbol-based 10.0 Introduction 10.5 Knowledge and Learning 10.1 A Framework for 10.6 Unsupervised Learning Symbol-based Learning 10.7 Reinforcement Learning 10.2 Version Space Search 10.8 Epilogue and 10.3 The

934 views • 54 slides
Genetics-based Machine Learning and Behaviour Based Robotics: A New Synthesis

Genetics-based Machine Learning and Behaviour Based Robotics: A New Synthesis

Genetics-based Machine Learning and Behaviour Based Robotics: A New Synthesis Genetics-based Machine Learning and Behavior Based Robotics: A New Synthesis, Marco Dorigo, Uwe Schnepf, IEEE Transactions on System, MAn, and Cybernetics,

348 views • 34 slides
Learning to Inflate Tom Rudelius IAS Based on 1810.05159/hep-th Outline Machine Learning

Learning to Inflate Tom Rudelius IAS Based on 1810.05159/hep-th Outline Machine Learning

Learning to Inflate Tom Rudelius IAS Based on 1810.05159/hep-th Outline Machine Learning and hep-th Introduction to Inflation Learning to Inflate Algorithm Results Possible Next Steps Machine Learning and hep-th Machine

709 views • 34 slides
Thinking about Genomics through a Machine Learning Lens: Basics of how several DNA- and RNA-based

Thinking about Genomics through a Machine Learning Lens: Basics of how several DNA- and RNA-based

Thinking about Genomics through a Machine Learning Lens: Basics of how several DNA- and RNA-based problems translate into commonly-studied areas of machine learning Faculty Lead Discussion (short version) 26 June 2018 GAUSSI Summer Retreat

323 views • 21 slides
INTRODUCTION TO MACHINE LEARNING Joseph C. Osborn CS 51A Spring 2020 Machine Learning is

INTRODUCTION TO MACHINE LEARNING Joseph C. Osborn CS 51A Spring 2020 Machine Learning is

INTRODUCTION TO MACHINE LEARNING Joseph C. Osborn CS 51A Spring 2020 Machine Learning is Machine learning is about predicting the future based on the past. -- Hal Daume III Machine Learning is Machine learning is about predicting

917 views • 59 slides
1 Model-Based Classification Model-Based Classification Model-based approach Build a

1 Model-Based Classification Model-Based Classification Model-based approach Build a

Machine Learning CSE 473: Artificial Intelligence Nave Bayes Up until now: how use a model to make optimal decisions Machine learning: how to acquire a model from data / experience Learning parameters (e.g. probabilities)

628 views • 8 slides
MCMC and Variational Inference for AutoEncoders Achille Thin 1 , Alain Durmus 2 , Eric Moulines 1 1

MCMC and Variational Inference for AutoEncoders Achille Thin 1 , Alain Durmus 2 , Eric Moulines 1 1

Introduction Deep Latent Generative Models (DLGMs) MetFlow and MetVAE: MCMC &amp; VI From classical to Flow-based MCMC Experiments MCMC and Variational Inference for AutoEncoders Achille Thin 1 , Alain Durmus 2 , Eric Moulines 1 1 Ecole

799 views • 67 slides
Practical Advice for Building Machine Learning Applications Machine Learning Based on lectures

Practical Advice for Building Machine Learning Applications Machine Learning Based on lectures

Practical Advice for Building Machine Learning Applications Machine Learning Based on lectures and papers by Andrew Ng, Pedro Domingos, Tom Mitchell and others 1 ML and the world Making ML work in the world Mostly experiential advice Also

863 views • 69 slides
Machine Learning over Complete Intersection Calabi-Yau Manifolds Workshop on Machine Learning

Machine Learning over Complete Intersection Calabi-Yau Manifolds Workshop on Machine Learning

Machine Learning over Complete Intersection Calabi-Yau Manifolds Workshop on Machine Learning Landscape ICTP, Trieste, Italy Challenger Mishra, ICMAT Madrid based on 1806.03121, and upcoming December 12, 2018 1. Physics Motivations 2.

655 views • 48 slides
An Exercise in An Exercise in Machine Learning Machine Learning

An Exercise in An Exercise in Machine Learning Machine Learning

An Exercise in An Exercise in Machine Learning Machine Learning http://www.cs.iastate.edu/~cs573x/bbsilab.html Machine Learning Software Preparing Data Building Classifiers Interpreting Results Machine Learning Software

496 views • 26 slides
PAC-learning, VC Dimension and Margin-based Bounds Machine Learning 10701/15781 Carlos

PAC-learning, VC Dimension and Margin-based Bounds Machine Learning 10701/15781 Carlos

More details: General: http://www.learning-with-kernels.org/ Example of more complex bounds: http://www.research.ibm.com/people/t/tzhang/papers/jmlr02_cover.ps.gz PAC-learning, VC Dimension and Margin-based Bounds Machine Learning

919 views • 59 slides
Optimizing multiplications with vector instructions Chitchanok Chuengsatiansup INRIA and ENS de

Optimizing multiplications with vector instructions Chitchanok Chuengsatiansup INRIA and ENS de

Optimizing multiplications with vector instructions Chitchanok Chuengsatiansup INRIA and ENS de Lyon 4 June 2018 Chitchanok Chuengsatiansup Optimizing multiplications with vector instructions 1 Introduction Current position: Postdoc (INRIA

1.32k views • 82 slides
Lecture 2 Combinational Logic Circuits Reference: Roth/John Text: Chapter 2 1 Combinational

Lecture 2 Combinational Logic Circuits Reference: Roth/John Text: Chapter 2 1 Combinational

Lecture 2 Combinational Logic Circuits Reference: Roth/John Text: Chapter 2 1 Combinational logic -- Behavior can be specified as concurrent signal assignments -- These model concurrent operation of hardware elements entity Gates is port

639 views • 59 slides
Karnaugh-Maps September 14, 2006 Typeset by Foil T EX What are Karnaugh Maps? A simpler

Karnaugh-Maps September 14, 2006 Typeset by Foil T EX What are Karnaugh Maps? A simpler

Karnaugh-Maps September 14, 2006 Typeset by Foil T EX What are Karnaugh Maps? A simpler way to handle most (but not all) jobs of manipulating logic functions. Typeset by Foil T EX 1 What are Karnaugh Maps? A simpler way to

811 views • 51 slides
CSEE 3827: Fundamentals of Computer Systems Lecture 4 &amp; 5 February 2 &amp; 4, 2009 Martha

CSEE 3827: Fundamentals of Computer Systems Lecture 4 &amp; 5 February 2 &amp; 4, 2009 Martha

CSEE 3827: Fundamentals of Computer Systems Lecture 4 &amp; 5 February 2 &amp; 4, 2009 Martha Kim martha@cs.columbia.edu Standard forms (redux) Product and sum terms Product term: logical AND of literals (e.g., XYZ) Sum term:

771 views • 40 slides
Minimizing Markov chains Beyond Bisimilarity* Giovanni Bacci, Giorgio Bacci, Kim G. Larsen , Radu

Minimizing Markov chains Beyond Bisimilarity* Giovanni Bacci, Giorgio Bacci, Kim G. Larsen , Radu

Minimizing Markov chains Beyond Bisimilarity* Giovanni Bacci, Giorgio Bacci, Kim G. Larsen , Radu Mardare Aalborg University, Denmark 22 April 2017 - Uppsala, Sweden SynCoP + PV 2017 (*) On the Metric-based Approximate Minimization of Markov

1.3k views • 87 slides
Lecture 7 Logistics HW2 due Wednesday --- Friday? Lab3 this week Lab3 this week

Lecture 7 Logistics HW2 due Wednesday --- Friday? Lab3 this week Lab3 this week

Lecture 7 Logistics HW2 due Wednesday --- Friday? Lab3 this week Lab3 this week 4/25 lecture canceled Last lecture Logic simplification Boolean cubes Karnaugh maps Today Today Continuing on K-maps with

536 views • 8 slides
Software Security Lucas Cordeiro Department of Computer Science lucas.cordeiro@manchester.ac.uk

Software Security Lucas Cordeiro Department of Computer Science lucas.cordeiro@manchester.ac.uk

Systems and Software Verification Laboratory Software Security Lucas Cordeiro Department of Computer Science lucas.cordeiro@manchester.ac.uk Career Summary 4 1 BSc/MSc in Engineering and Lecturer Career Summary 2 1 BSc/MSc in MSc in

1.95k views • 165 slides
Seiberg-Witten Theory and AGT Relation Tohru Eguchi We consider N = 2 supersymmetric gauge

Seiberg-Witten Theory and AGT Relation Tohru Eguchi We consider N = 2 supersymmetric gauge

Seiberg-Witten Theory and AGT Relation Tohru Eguchi We consider N = 2 supersymmetric gauge theories in 4- dimensions and study the case when the theory possesses the conformal invariance. Simplest example of a conformal invariant theory:

904 views • 38 slides

More recommend