Predictive Processing in Planning:
|
|
- Vernon Williamson
- 5 years ago
- Views:
Transcription
1 Predictive Processing in Planning: Choice Behavior as Active Bayesian Inference Philipp Schwartenbeck Wellcome Trust Centre for Human Neuroimaging, UCL The Promise of Predictive Processing: A Critical Evaluation of its Prospects Tufts University 2018
2 Inferring the latent structure of the world v. Helmholtz, 1867 Active inference: to make sense of the world, we need to infer its latent structure (hidden causes) This is not only relevant for perception, but also for adaptive behavior!
3 What is active inference? world s a t A agent sƹ Bayesian Brain, Good Regulator Theorem (Conant & Ashby, 1947).. o t O Central idea: Action fulfils prior expectations = preferences o F = ln P(o t ) D[Q(s t ) P(s t o t )] Qs ( ) Evidence Divergence Free energy Taken from Friston, 2017 minimise surprise predictive processing Action Changing sensations to maximise evidence Perception Changing beliefs to minimise divergence
4 Outline 1. Active Inference: basic building blocks i. Markov Decision Processes ii. Generative models iii. Information theory iv. Variational Bayes/free energy world s a t A o t O agent sƹ 2. Active inference based on a discrete Markovian generative model 3. Some applications and predictions
5 Active inference: basic building blocks
6 I. Markov Decision Processes Theoretical framework for decision-making: partially observable Markov decision processes (POMDP) Key ingredients: 1,, T discrete time-steps Hidden states s t 1 s s t t 1
7 I. Markov Decision Processes Theoretical framework for decision-making: partially observable Markov decision processes (POMDP) Key ingredients: 1,, T discrete time-steps s t vs. P o t s t ) not trivial o 1 o 2 o3
8 I. Markov Decision Processes Theoretical framework for decision-making: partially observable Markov decision processes (POMDP) Key ingredients: 1,, T discrete time-steps a t 1 s t+1 P o t s t ) not trivial a t 2 a t 1 Hidden states s t 2 t 1 s t+1 s s t a t a t 2 3 s t+1 Markov-property: P s t+1 s t, a t = P s t+1 s t, a t, s t 1, a t 1, s t 2, a t 2,
9 II. Generative models and hidden states Generative models are probabilistic mappings based on a likelihood function and a prior density: p y, θ m = p y θ, m p θ, m [ p o t, s t = p o t s t p s t ] Allows to generate data by sampling from prior and plugging into likelihood Call θ/s t hidden state or latent variable. Perform inference on hidden states by applying Bayes rule ( model inversion ): p θ y, m = p y θ, m p θ m p(y m) Active inference emphasizes that agents build generative models on the world, and use them to perform inference on perception and action.
10 III. Information theory Information theory quantifies the information content of a signal Unlikely events are more informative than likely events This can be quantified as the self-information or surprise of a signal (Shannon, 1948): I y = logp(y m) Expected value of surprise is called (Shannon) entropy: H y = E I y Expected amount of information of an event (important later!) Also: measure of uncertainty = p y i m log p(y i m) i Active inference is heavily grounded in information theory, and argues that it has a lot to say about how brains work and how we behave.
11 IV. Variational Bayes Exact inference is generally intractable p θ y, m = p y θ, m p(θ m) p(y m) = p y θ, m p(θ m) p y θ, m p(θ m) Variational Bayes allows to cast an inference problem (difficult) as a bound optimization problem (easier) (cf., Beal, 2003; Bogacz 2017) Aim: approximate model evidence (denominator in Bayes) (Negative) variational free energy provides a lower bound on model evidence General formula: F = log p y m D KL q θ, p θ y, m Accuracy Complexity Note: negative log-model evidence is the same as information-theoretic surprise log P y m! log p y m = F D KL q θ, p θ y, m
12 Active inference: building blocks I. Choice behavior is modeled as partially observable Markov Decision Process Highlights importance of inferring hidden states and adequate choices II. III. Inference on hidden states and choices is based on an agent s generative model of the world Invert models via Bayesian inference Information theory allows to quantify how informative/expected an observation is Surprise = negative (log-) model evidence! IV. Inference is approximated based on (negative) variational free energy Approximates model evidence (surprise!)
13 Active Inference: model structure
14 ǁ ǁ A Markovian generative model p y, θ = p y θ p θ P o, s, ǁ π, γ = P o s P s π P(π γ)p(γ) P o s ǁ = P o 0 s 0 P o 1 s 1 P o t s t P o t s t =: A Likelihood P sǁ π = P s t s t 1, π P s 1 s 0, π P(s 0 ) P s t+1 s t, π =: B(u = π(t)) P s 0 P o t =: D =: σ(c) Empirical priors hidden states d Precision C P(π γ) = σ( γ Q) control states Policy Q = E[ln P o t s t ln Q o t π + ln P(o t )] t D B P A = Dir(a) P B = Dir(b) P D = Dir(d) P γ = Γ(1, β) Full priors Hidden states s t 1 st st 1 A Q s, ǁ π, A, B, D, γ = Q s 1 π Q s T π Q π Q A Q B Q D Q(γ) Q s t π = Cat(s t ) Q(π) = Cat(π) ot 1 o t ot 1 Q(A) = Dir(a) Q(B) = Dir(b) Q(D) = Dir(d) Q(γ) = Γ(1, β) Approximate posterior
15 A-matrix: observation model P o, s, ǁ π, γ = P o sǁ P sǁ π P(π γ)p(γ) Hidden states s t 1 st st 1 P o s ǁ = P o 0 s 0 P o 1 s 1 P o t s t A P o t s t =: A Likelihood o o t 1 t o t 1 P
16 B-Matrix: Markovian transition probabilities P o, s, ǁ π, γ = P o sǁ P sǁ π P(π γ)p(γ) Policy P sǁ π = P s t s t 1, π P s 1 s 0, π P(s 0 ) P s t+1 s t, π =: B(u = π(t)) P s 0 P o t =: D =: σ(c) Empirical priors hidden states Hidden states s t 1 st st 1 B P, gamble
17 Prior expectations over outcomes (c-vector) P o, s, ǁ π, γ = P o sǁ P sǁ π P(π γ)p(γ) P sǁ π = P s t s t 1, π P s 1 s 0, π P(s 0 ) P s t+1 s t, π =: B(u = π(t)) P s 0 P o t =: D =: σ(c) Empirical priors hidden states d D Policy C Hidden states s t 1 s s t t 1
18 Precision γ P o, s, ǁ π, γ = P o sǁ P sǁ π P(π γ)p(γ) P(π γ) = σ( γ Q) control states Q = E [H[P o t s t ]] + H Q o t π + E[lnP(o t )] t P A = Dir(a) P D = Dir(d) P B = Dir(b) P γ = Γ(1, β) Full priors Precision Policy gamble vs. don t gamble?
19 Precision and Uncertainty Precision encoded by neuromodulators? ACh? NA? DA? Schwartenbeck et al., 2015 Parr & Friston, 2017
20 Inference Belief updating Maths is not trivial but only has to be done once! (By mathematicians, evolution, ) Then implement resulting update equations. Perception Policy selection Precision π sƹ t = σ(log A o t + log B t 1 sƹ t 1 ) π = σ( γ Q) መβ = β π Q s F π F γ F Q Derive via variational free energy with respect to hidden states x = {s t, π, β}: Q x = arg min Q(x) F(o, x, Q) P(x o) F o, x, Q = E Q st,π,β ln P o, s t, π, β m E Q st,π,β [ln Q s t, π, β ] π = s t (ln sƹ t ln A o t ln B t 1 sƹ t 1 ) + cf., Friston et al., 2015, 2017; Bogacz, 2017
21 Inference: a closer (more intuitive) look Belief updating Perception π sƹ t = σ(log A o t + log B t 1 sƹ t 1 ) Likely hidden states given what I observe Based on what I believe about the previous state Policy selection π = σ( γ Q) Confidence Value of policies Q Precision መβ = β π Q Prior on precision Value of inferred policy cf., Friston et al., 2015, 2017
22 Beyond (active) inference: learning Belief updating Perception Policy selection Precision sƹ t = σ(log A o t + log B π sƹ t 1 ) π = σ( γ Q) መβ = β π Q???? Q A = ψ a ψ(a 0 ) a = a + o t D = ψ d ψ(d 0 ) d = d + s t Learning cf., Friston et al., 2016
23 Policy selection as predictive processing Policies become valuable if they Maximize reward/utility Keep options open Allow us to minimize uncertainty about the world Utilities or preferences are defined as log-expectations over outcomes: P(o t ) log P(o t ) Fulfilling these preferences maximizes model evidence log P o t (i.e., minimizes surprise log P(o t ))! This can be approximated with variational free energy!
24 Policy selection as predictive processing Value of policies Q(π) defined as negative free energy: Q s, ǁ π, A, B, D, γ = Q s 1 π Q s T π Q π Q A Q B Q D Q(γ) Approximate posterior Q(π) = E[H[P o t s t ]] D KL [Q o t π P(o t )] t Friston et al., 2015 Q(π) = E [H[P o t s t ]] + H Q o t π + E[lnP(o t )] t Expected uncertainty (H = 0 o t = s t ) Entropy over outcomes Expected utility Epistemic or intrinsic value Extrinsic value Friston et al., 2015; 2017; Parr & Friston, 2017
25 Some Applications and Predictions
26 Inference: a closer look on policy selection Imagine you are a Tennis player and your opponent serves You want to find a position where Expected utility: E[lnP(o t )]...there is a high chance of getting the ball, Entropy over outcomes: H Q o t π... you have a good chance of getting a ball that was unexpected,...and where you can minimize uncertainty about the character of your opponent. Ambiguity: E[H P o t s t ]
27 Policy selection: expected utility vs. entropy over outcomes Expected Utility E lnp o t = o t π U t Entropy over outcomes Expected Utility H Q o t π + E lnp o t = o t π U t o t π log o t π Schwartenbeck et al., 2015
28 Policy selection: foraging Friston,, FitzGerald & Pezzulo, 2015 foraging trials Schwartenbeck & Friston, 2016
29 Policy selection: curiosity and exploration P( o, s, ǁ A, π, γ) Q π = E Q ln P o t E Q [D KL [P(s t o t, π) Q s t π + D KL [P(A o t, π) Q A π ] cf., Parr & Friston, 2017 Schwartenbeck et al. in prep Expected update Exploration? {, }? Exploitation
30 p(risky) p(no reward risky option) p(risky) p(no reward risky option) p(risky) Policy selection: curiosity and exploration Active learning: Goal-directed uncertainty reduction in beginning of experiment p(reward risky option) trials Random learning: Random uncertainty reduction p(reward risky option) trials No learning: No uncertainty reduction trials Schwartenbeck et al. in prep
31 State estimation: Planning as inference s π t = σ(log A o t + log B π sƹ π t 1 ) sƹ t = State estimation as Bayesian model averaging of likely states under different policies: i π i s t π p s t = p s t π i p(π i ) i This predicts an optimism bias for estimates of current (and future states) This accounts for trade-offs and hybrid models of different modes of behaviour, e.g. model-free vs. model-based behaviour Predicts accuracy-complexity trade-off: FitzGerald, Dolan & Friston, 2014
32 Building models of the world world s a t A agent sƹ o t O???? Q FitzGerald, Dolan & Friston, 2014
33 Summary Active inference offers probabilistic treatment of planning and decision-making Highlights importance of developing a (generative) model to perform inference Introduce predictive processing to decision-making: Choices fulfil preferences minimise surprise maximise model evidence Approximated by variational free energy Approximate inference takes place based on variational updates of the current state, policy and confidence Applies the same currency to information-gain and reward, such that agents Maximize utility Keep options open Minimize uncertainty ( goal-directed exploration)
34 Thank you!
Computational Biology Lecture #3: Probability and Statistics. Bud Mishra Professor of Computer Science, Mathematics, & Cell Biology Sept
Computational Biology Lecture #3: Probability and Statistics Bud Mishra Professor of Computer Science, Mathematics, & Cell Biology Sept 26 2005 L2-1 Basic Probabilities L2-2 1 Random Variables L2-3 Examples
More informationMachine Learning and Bayesian Inference. Unsupervised learning. Can we find regularity in data without the aid of labels?
Machine Learning and Bayesian Inference Dr Sean Holden Computer Laboratory, Room FC6 Telephone extension 6372 Email: sbh11@cl.cam.ac.uk www.cl.cam.ac.uk/ sbh11/ Unsupervised learning Can we find regularity
More informationBayesian Inference Course, WTCN, UCL, March 2013
Bayesian Course, WTCN, UCL, March 2013 Shannon (1948) asked how much information is received when we observe a specific value of the variable x? If an unlikely event occurs then one would expect the information
More informationG8325: Variational Bayes
G8325: Variational Bayes Vincent Dorie Columbia University Wednesday, November 2nd, 2011 bridge Variational University Bayes Press 2003. On-screen viewing permitted. Printing not permitted. http://www.c
More informationBelief and Desire: On Information and its Value
Belief and Desire: On Information and its Value Ariel Caticha Department of Physics University at Albany SUNY ariel@albany.edu Info-Metrics Institute 04/26/2013 1 Part 1: Belief 2 What is information?
More informationIntroduction to Mobile Robotics Probabilistic Robotics
Introduction to Mobile Robotics Probabilistic Robotics Wolfram Burgard 1 Probabilistic Robotics Key idea: Explicit representation of uncertainty (using the calculus of probability theory) Perception Action
More informationStatistical learning. Chapter 20, Sections 1 4 1
Statistical learning Chapter 20, Sections 1 4 Chapter 20, Sections 1 4 1 Outline Bayesian learning Maximum a posteriori and maximum likelihood learning Bayes net learning ML parameter learning with complete
More informationUncertainty, precision, prediction errors and their relevance to computational psychiatry
Uncertainty, precision, prediction errors and their relevance to computational psychiatry Christoph Mathys Wellcome Trust Centre for Neuroimaging at UCL, London, UK Max Planck UCL Centre for Computational
More informationCS 630 Basic Probability and Information Theory. Tim Campbell
CS 630 Basic Probability and Information Theory Tim Campbell 21 January 2003 Probability Theory Probability Theory is the study of how best to predict outcomes of events. An experiment (or trial or event)
More informationComputational Perception. Bayesian Inference
Computational Perception 15-485/785 January 24, 2008 Bayesian Inference The process of probabilistic inference 1. define model of problem 2. derive posterior distributions and estimators 3. estimate parameters
More informationNotes from Week 9: Multi-Armed Bandit Problems II. 1 Information-theoretic lower bounds for multiarmed
CS 683 Learning, Games, and Electronic Markets Spring 007 Notes from Week 9: Multi-Armed Bandit Problems II Instructor: Robert Kleinberg 6-30 Mar 007 1 Information-theoretic lower bounds for multiarmed
More informationActive Inference: A Process Theory
ARTICLE Communicated by Ollie Hulme Active Inference: A Process Theory Karl Friston k.friston@ucl.ac.uk Wellcome Trust Centre for Neuroimaging, UCL, London WC1N 3BG, U.K. Thomas FitzGerald thomas.fitzgerald@ucl.ac.uk
More informationIntrinsic and Extrinsic Motivation
Intrinsic and Extrinsic Motivation Roland Bénabou Jean Tirole. Review of Economic Studies 2003 Bénabou and Tirole Intrinsic and Extrinsic Motivation 1 / 30 Motivation Should a child be rewarded for passing
More informationProbabilistic Reasoning in Deep Learning
Probabilistic Reasoning in Deep Learning Dr Konstantina Palla, PhD palla@stats.ox.ac.uk September 2017 Deep Learning Indaba, Johannesburgh Konstantina Palla 1 / 39 OVERVIEW OF THE TALK Basics of Bayesian
More informationVariational Methods in Bayesian Deconvolution
PHYSTAT, SLAC, Stanford, California, September 8-, Variational Methods in Bayesian Deconvolution K. Zarb Adami Cavendish Laboratory, University of Cambridge, UK This paper gives an introduction to the
More informationBioinformatics: Biology X
Bud Mishra Room 1002, 715 Broadway, Courant Institute, NYU, New York, USA Model Building/Checking, Reverse Engineering, Causality Outline 1 Bayesian Interpretation of Probabilities 2 Where (or of what)
More informationPlanning by Probabilistic Inference
Planning by Probabilistic Inference Hagai Attias Microsoft Research 1 Microsoft Way Redmond, WA 98052 Abstract This paper presents and demonstrates a new approach to the problem of planning under uncertainty.
More informationStatistical learning. Chapter 20, Sections 1 3 1
Statistical learning Chapter 20, Sections 1 3 Chapter 20, Sections 1 3 1 Outline Bayesian learning Maximum a posteriori and maximum likelihood learning Bayes net learning ML parameter learning with complete
More informationIntroduction to Systems Analysis and Decision Making Prepared by: Jakub Tomczak
Introduction to Systems Analysis and Decision Making Prepared by: Jakub Tomczak 1 Introduction. Random variables During the course we are interested in reasoning about considered phenomenon. In other words,
More informationDecision Theory: Q-Learning
Decision Theory: Q-Learning CPSC 322 Decision Theory 5 Textbook 12.5 Decision Theory: Q-Learning CPSC 322 Decision Theory 5, Slide 1 Lecture Overview 1 Recap 2 Asynchronous Value Iteration 3 Q-Learning
More informationPosterior Regularization
Posterior Regularization 1 Introduction One of the key challenges in probabilistic structured learning, is the intractability of the posterior distribution, for fast inference. There are numerous methods
More information13: Variational inference II
10-708: Probabilistic Graphical Models, Spring 2015 13: Variational inference II Lecturer: Eric P. Xing Scribes: Ronghuo Zheng, Zhiting Hu, Yuntian Deng 1 Introduction We started to talk about variational
More informationHGF Workshop. Christoph Mathys. Wellcome Trust Centre for Neuroimaging at UCL, Max Planck UCL Centre for Computational Psychiatry and Ageing Research
HGF Workshop Christoph Mathys Wellcome Trust Centre for Neuroimaging at UCL, Max Planck UCL Centre for Computational Psychiatry and Ageing Research Monash University, March 3, 2015 Systems theory places
More informationParametric Models. Dr. Shuang LIANG. School of Software Engineering TongJi University Fall, 2012
Parametric Models Dr. Shuang LIANG School of Software Engineering TongJi University Fall, 2012 Today s Topics Maximum Likelihood Estimation Bayesian Density Estimation Today s Topics Maximum Likelihood
More informationCS 4649/7649 Robot Intelligence: Planning
CS 4649/7649 Robot Intelligence: Planning Probability Primer Sungmoon Joo School of Interactive Computing College of Computing Georgia Institute of Technology S. Joo (sungmoon.joo@cc.gatech.edu) 1 *Slides
More informationBayesian Inference and MCMC
Bayesian Inference and MCMC Aryan Arbabi Partly based on MCMC slides from CSC412 Fall 2018 1 / 18 Bayesian Inference - Motivation Consider we have a data set D = {x 1,..., x n }. E.g each x i can be the
More informationLecture 2: From Linear Regression to Kalman Filter and Beyond
Lecture 2: From Linear Regression to Kalman Filter and Beyond Department of Biomedical Engineering and Computational Science Aalto University January 26, 2012 Contents 1 Batch and Recursive Estimation
More informationConditional probabilities and graphical models
Conditional probabilities and graphical models Thomas Mailund Bioinformatics Research Centre (BiRC), Aarhus University Probability theory allows us to describe uncertainty in the processes we model within
More informationMathematical Formulation of Our Example
Mathematical Formulation of Our Example We define two binary random variables: open and, where is light on or light off. Our question is: What is? Computer Vision 1 Combining Evidence Suppose our robot
More informationA Bayesian model for event-based trust
A Bayesian model for event-based trust Elements of a foundation for computational trust Vladimiro Sassone ECS, University of Southampton joint work K. Krukow and M. Nielsen Oxford, 9 March 2007 V. Sassone
More informationBayesian Learning. HT2015: SC4 Statistical Data Mining and Machine Learning. Maximum Likelihood Principle. The Bayesian Learning Framework
HT5: SC4 Statistical Data Mining and Machine Learning Dino Sejdinovic Department of Statistics Oxford http://www.stats.ox.ac.uk/~sejdinov/sdmml.html Maximum Likelihood Principle A generative model for
More informationCSC321 Lecture 18: Learning Probabilistic Models
CSC321 Lecture 18: Learning Probabilistic Models Roger Grosse Roger Grosse CSC321 Lecture 18: Learning Probabilistic Models 1 / 25 Overview So far in this course: mainly supervised learning Language modeling
More informationBayesian Networks BY: MOHAMAD ALSABBAGH
Bayesian Networks BY: MOHAMAD ALSABBAGH Outlines Introduction Bayes Rule Bayesian Networks (BN) Representation Size of a Bayesian Network Inference via BN BN Learning Dynamic BN Introduction Conditional
More informationVIME: Variational Information Maximizing Exploration
1 VIME: Variational Information Maximizing Exploration Rein Houthooft, Xi Chen, Yan Duan, John Schulman, Filip De Turck, Pieter Abbeel Reviewed by Zhao Song January 13, 2017 1 Exploration in Reinforcement
More informationLearning in Bayesian Networks
Learning in Bayesian Networks Florian Markowetz Max-Planck-Institute for Molecular Genetics Computational Molecular Biology Berlin Berlin: 20.06.2002 1 Overview 1. Bayesian Networks Stochastic Networks
More informationSTA 4273H: Sta-s-cal Machine Learning
STA 4273H: Sta-s-cal Machine Learning Russ Salakhutdinov Department of Computer Science! Department of Statistical Sciences! rsalakhu@cs.toronto.edu! h0p://www.cs.utoronto.ca/~rsalakhu/ Lecture 2 In our
More informationProbabilistic numerics for deep learning
Presenter: Shijia Wang Department of Engineering Science, University of Oxford rning (RLSS) Summer School, Montreal 2017 Outline 1 Introduction Probabilistic Numerics 2 Components Probabilistic modeling
More informationPartially Observable Markov Decision Processes (POMDPs) Pieter Abbeel UC Berkeley EECS
Partially Observable Markov Decision Processes (POMDPs) Pieter Abbeel UC Berkeley EECS Many slides adapted from Jur van den Berg Outline POMDPs Separation Principle / Certainty Equivalence Locally Optimal
More informationDialogue as a Decision Making Process
Dialogue as a Decision Making Process Nicholas Roy Challenges of Autonomy in the Real World Wide range of sensors Noisy sensors World dynamics Adaptability Incomplete information Robustness under uncertainty
More informationThe Naïve Bayes Classifier. Machine Learning Fall 2017
The Naïve Bayes Classifier Machine Learning Fall 2017 1 Today s lecture The naïve Bayes Classifier Learning the naïve Bayes Classifier Practical concerns 2 Today s lecture The naïve Bayes Classifier Learning
More informationBayesian inference. Fredrik Ronquist and Peter Beerli. October 3, 2007
Bayesian inference Fredrik Ronquist and Peter Beerli October 3, 2007 1 Introduction The last few decades has seen a growing interest in Bayesian inference, an alternative approach to statistical inference.
More informationDensity Estimation. Seungjin Choi
Density Estimation Seungjin Choi Department of Computer Science and Engineering Pohang University of Science and Technology 77 Cheongam-ro, Nam-gu, Pohang 37673, Korea seungjin@postech.ac.kr http://mlg.postech.ac.kr/
More informationLearning with Probabilities
Learning with Probabilities CS194-10 Fall 2011 Lecture 15 CS194-10 Fall 2011 Lecture 15 1 Outline Bayesian learning eliminates arbitrary loss functions and regularizers facilitates incorporation of prior
More informationPATTERN RECOGNITION AND MACHINE LEARNING
PATTERN RECOGNITION AND MACHINE LEARNING Chapter 1. Introduction Shuai Huang April 21, 2014 Outline 1 What is Machine Learning? 2 Curve Fitting 3 Probability Theory 4 Model Selection 5 The curse of dimensionality
More informationDiscrete Mathematics and Probability Theory Fall 2015 Lecture 21
CS 70 Discrete Mathematics and Probability Theory Fall 205 Lecture 2 Inference In this note we revisit the problem of inference: Given some data or observations from the world, what can we infer about
More informationIntroduction: MLE, MAP, Bayesian reasoning (28/8/13)
STA561: Probabilistic machine learning Introduction: MLE, MAP, Bayesian reasoning (28/8/13) Lecturer: Barbara Engelhardt Scribes: K. Ulrich, J. Subramanian, N. Raval, J. O Hollaren 1 Classifiers In this
More informationLecture : Probabilistic Machine Learning
Lecture : Probabilistic Machine Learning Riashat Islam Reasoning and Learning Lab McGill University September 11, 2018 ML : Many Methods with Many Links Modelling Views of Machine Learning Machine Learning
More informationVariational Decoding for Statistical Machine Translation
Variational Decoding for Statistical Machine Translation Zhifei Li, Jason Eisner, and Sanjeev Khudanpur Center for Language and Speech Processing Computer Science Department Johns Hopkins University 1
More informationA.I. in health informatics lecture 2 clinical reasoning & probabilistic inference, I. kevin small & byron wallace
A.I. in health informatics lecture 2 clinical reasoning & probabilistic inference, I kevin small & byron wallace today a review of probability random variables, maximum likelihood, etc. crucial for clinical
More informationCS 188: Artificial Intelligence Spring Today
CS 188: Artificial Intelligence Spring 2006 Lecture 9: Naïve Bayes 2/14/2006 Dan Klein UC Berkeley Many slides from either Stuart Russell or Andrew Moore Bayes rule Today Expectations and utilities Naïve
More informationVariational Inference in TensorFlow. Danijar Hafner Stanford CS University College London, Google Brain
Variational Inference in TensorFlow Danijar Hafner Stanford CS 20 2018-02-16 University College London, Google Brain Outline Variational Inference Tensorflow Distributions VAE in TensorFlow Variational
More informationModeling Environment
Topic Model Modeling Environment What does it mean to understand/ your environment? Ability to predict Two approaches to ing environment of words and text Latent Semantic Analysis (LSA) Topic Model LSA
More informationStatistical learning. Chapter 20, Sections 1 3 1
Statistical learning Chapter 20, Sections 1 3 Chapter 20, Sections 1 3 1 Outline Bayesian learning Maximum a posteriori and maximum likelihood learning Bayes net learning ML parameter learning with complete
More informationBayesian modeling of learning and decision-making in uncertain environments
Bayesian modeling of learning and decision-making in uncertain environments Christoph Mathys Max Planck UCL Centre for Computational Psychiatry and Ageing Research, London, UK Wellcome Trust Centre for
More informationIEOR E4570: Machine Learning for OR&FE Spring 2015 c 2015 by Martin Haugh. The EM Algorithm
IEOR E4570: Machine Learning for OR&FE Spring 205 c 205 by Martin Haugh The EM Algorithm The EM algorithm is used for obtaining maximum likelihood estimates of parameters when some of the data is missing.
More informationA prior distribution over model space p(m) (or hypothesis space ) can be updated to a posterior distribution after observing data y.
June 2nd 2011 A prior distribution over model space p(m) (or hypothesis space ) can be updated to a posterior distribution after observing data y. This is implemented using Bayes rule p(m y) = p(y m)p(m)
More informationExpectation Propagation Algorithm
Expectation Propagation Algorithm 1 Shuang Wang School of Electrical and Computer Engineering University of Oklahoma, Tulsa, OK, 74135 Email: {shuangwang}@ou.edu This note contains three parts. First,
More informationProbability and Information Theory. Sargur N. Srihari
Probability and Information Theory Sargur N. srihari@cedar.buffalo.edu 1 Topics in Probability and Information Theory Overview 1. Why Probability? 2. Random Variables 3. Probability Distributions 4. Marginal
More informationRobust Monte Carlo Methods for Sequential Planning and Decision Making
Robust Monte Carlo Methods for Sequential Planning and Decision Making Sue Zheng, Jason Pacheco, & John Fisher Sensing, Learning, & Inference Group Computer Science & Artificial Intelligence Laboratory
More informationGrundlagen der Künstlichen Intelligenz
Grundlagen der Künstlichen Intelligenz Uncertainty & Probabilities & Bandits Daniel Hennes 16.11.2017 (WS 2017/18) University Stuttgart - IPVS - Machine Learning & Robotics 1 Today Uncertainty Probability
More information6 Reinforcement Learning
6 Reinforcement Learning As discussed above, a basic form of supervised learning is function approximation, relating input vectors to output vectors, or, more generally, finding density functions p(y,
More informationIntroduction to Machine Learning
Introduction to Machine Learning Brown University CSCI 1950-F, Spring 2012 Prof. Erik Sudderth Lecture 25: Markov Chain Monte Carlo (MCMC) Course Review and Advanced Topics Many figures courtesy Kevin
More informationLecture 6: Graphical Models: Learning
Lecture 6: Graphical Models: Learning 4F13: Machine Learning Zoubin Ghahramani and Carl Edward Rasmussen Department of Engineering, University of Cambridge February 3rd, 2010 Ghahramani & Rasmussen (CUED)
More informationParametric Unsupervised Learning Expectation Maximization (EM) Lecture 20.a
Parametric Unsupervised Learning Expectation Maximization (EM) Lecture 20.a Some slides are due to Christopher Bishop Limitations of K-means Hard assignments of data points to clusters small shift of a
More informationBayesian Methods. David S. Rosenberg. New York University. March 20, 2018
Bayesian Methods David S. Rosenberg New York University March 20, 2018 David S. Rosenberg (New York University) DS-GA 1003 / CSCI-GA 2567 March 20, 2018 1 / 38 Contents 1 Classical Statistics 2 Bayesian
More informationBasic Probabilistic Reasoning SEG
Basic Probabilistic Reasoning SEG 7450 1 Introduction Reasoning under uncertainty using probability theory Dealing with uncertainty is one of the main advantages of an expert system over a simple decision
More informationVariational Inference via Stochastic Backpropagation
Variational Inference via Stochastic Backpropagation Kai Fan February 27, 2016 Preliminaries Stochastic Backpropagation Variational Auto-Encoding Related Work Summary Outline Preliminaries Stochastic Backpropagation
More informationAnother Walkthrough of Variational Bayes. Bevan Jones Machine Learning Reading Group Macquarie University
Another Walkthrough of Variational Bayes Bevan Jones Machine Learning Reading Group Macquarie University 2 Variational Bayes? Bayes Bayes Theorem But the integral is intractable! Sampling Gibbs, Metropolis
More informationPMR Learning as Inference
Outline PMR Learning as Inference Probabilistic Modelling and Reasoning Amos Storkey Modelling 2 The Exponential Family 3 Bayesian Sets School of Informatics, University of Edinburgh Amos Storkey PMR Learning
More information1 Introduction 2. 4 Q-Learning The Q-value The Temporal Difference The whole Q-Learning process... 5
Table of contents 1 Introduction 2 2 Markov Decision Processes 2 3 Future Cumulative Reward 3 4 Q-Learning 4 4.1 The Q-value.............................................. 4 4.2 The Temporal Difference.......................................
More informationCrowdsourcing & Optimal Budget Allocation in Crowd Labeling
Crowdsourcing & Optimal Budget Allocation in Crowd Labeling Madhav Mohandas, Richard Zhu, Vincent Zhuang May 5, 2016 Table of Contents 1. Intro to Crowdsourcing 2. The Problem 3. Knowledge Gradient Algorithm
More informationBayesian Inference. Introduction
Bayesian Inference Introduction The frequentist approach to inference holds that probabilities are intrinsicially tied (unsurprisingly) to frequencies. This interpretation is actually quite natural. What,
More informationMachine Learning Summer School
Machine Learning Summer School Lecture 3: Learning parameters and structure Zoubin Ghahramani zoubin@eng.cam.ac.uk http://learning.eng.cam.ac.uk/zoubin/ Department of Engineering University of Cambridge,
More informationJohn Ashburner. Wellcome Trust Centre for Neuroimaging, UCL Institute of Neurology, 12 Queen Square, London WC1N 3BG, UK.
FOR MEDICAL IMAGING John Ashburner Wellcome Trust Centre for Neuroimaging, UCL Institute of Neurology, 12 Queen Square, London WC1N 3BG, UK. PIPELINES V MODELS PROBABILITY THEORY MEDICAL IMAGE COMPUTING
More informationPILCO: A Model-Based and Data-Efficient Approach to Policy Search
PILCO: A Model-Based and Data-Efficient Approach to Policy Search (M.P. Deisenroth and C.E. Rasmussen) CSC2541 November 4, 2016 PILCO Graphical Model PILCO Probabilistic Inference for Learning COntrol
More informationREINFORCE Framework for Stochastic Policy Optimization and its use in Deep Learning
REINFORCE Framework for Stochastic Policy Optimization and its use in Deep Learning Ronen Tamari The Hebrew University of Jerusalem Advanced Seminar in Deep Learning (#67679) February 28, 2016 Ronen Tamari
More informationTwo Useful Bounds for Variational Inference
Two Useful Bounds for Variational Inference John Paisley Department of Computer Science Princeton University, Princeton, NJ jpaisley@princeton.edu Abstract We review and derive two lower bounds on the
More informationBayesian Learning. Artificial Intelligence Programming. 15-0: Learning vs. Deduction
15-0: Learning vs. Deduction Artificial Intelligence Programming Bayesian Learning Chris Brooks Department of Computer Science University of San Francisco So far, we ve seen two types of reasoning: Deductive
More informationLecture 13 : Variational Inference: Mean Field Approximation
10-708: Probabilistic Graphical Models 10-708, Spring 2017 Lecture 13 : Variational Inference: Mean Field Approximation Lecturer: Willie Neiswanger Scribes: Xupeng Tong, Minxing Liu 1 Problem Setup 1.1
More informationCOMS 4721: Machine Learning for Data Science Lecture 16, 3/28/2017
COMS 4721: Machine Learning for Data Science Lecture 16, 3/28/2017 Prof. John Paisley Department of Electrical Engineering & Data Science Institute Columbia University SOFT CLUSTERING VS HARD CLUSTERING
More informationProbabilistic Graphical Models
Probabilistic Graphical Models Introduction. Basic Probability and Bayes Volkan Cevher, Matthias Seeger Ecole Polytechnique Fédérale de Lausanne 26/9/2011 (EPFL) Graphical Models 26/9/2011 1 / 28 Outline
More information(1) Introduction to Bayesian statistics
Spring, 2018 A motivating example Student 1 will write down a number and then flip a coin If the flip is heads, they will honestly tell student 2 if the number is even or odd If the flip is tails, they
More informationMarkov localization uses an explicit, discrete representation for the probability of all position in the state space.
Markov Kalman Filter Localization Markov localization localization starting from any unknown position recovers from ambiguous situation. However, to update the probability of all positions within the whole
More informationIntroduction to Bayesian inference
Introduction to Bayesian inference Thomas Alexander Brouwer University of Cambridge tab43@cam.ac.uk 17 November 2015 Probabilistic models Describe how data was generated using probability distributions
More informationReinforcement learning
Reinforcement learning Stuart Russell, UC Berkeley Stuart Russell, UC Berkeley 1 Outline Sequential decision making Dynamic programming algorithms Reinforcement learning algorithms temporal difference
More information13 : Variational Inference: Loopy Belief Propagation and Mean Field
10-708: Probabilistic Graphical Models 10-708, Spring 2012 13 : Variational Inference: Loopy Belief Propagation and Mean Field Lecturer: Eric P. Xing Scribes: Peter Schulam and William Wang 1 Introduction
More informationOutline. CSE 573: Artificial Intelligence Autumn Agent. Partial Observability. Markov Decision Process (MDP) 10/31/2012
CSE 573: Artificial Intelligence Autumn 2012 Reasoning about Uncertainty & Hidden Markov Models Daniel Weld Many slides adapted from Dan Klein, Stuart Russell, Andrew Moore & Luke Zettlemoyer 1 Outline
More informationNaïve Bayes classification
Naïve Bayes classification 1 Probability theory Random variable: a variable whose possible values are numerical outcomes of a random phenomenon. Examples: A person s height, the outcome of a coin toss
More information2. I will provide two examples to contrast the two schools. Hopefully, in the end you will fall in love with the Bayesian approach.
Bayesian Statistics 1. In the new era of big data, machine learning and artificial intelligence, it is important for students to know the vocabulary of Bayesian statistics, which has been competing with
More informationNPFL108 Bayesian inference. Introduction. Filip Jurčíček. Institute of Formal and Applied Linguistics Charles University in Prague Czech Republic
NPFL108 Bayesian inference Introduction Filip Jurčíček Institute of Formal and Applied Linguistics Charles University in Prague Czech Republic Home page: http://ufal.mff.cuni.cz/~jurcicek Version: 21/02/2014
More informationStratégies bayésiennes et fréquentistes dans un modèle de bandit
Stratégies bayésiennes et fréquentistes dans un modèle de bandit thèse effectuée à Telecom ParisTech, co-dirigée par Olivier Cappé, Aurélien Garivier et Rémi Munos Journées MAS, Grenoble, 30 août 2016
More informationA reinforcement learning scheme for a multi-agent card game with Monte Carlo state estimation
A reinforcement learning scheme for a multi-agent card game with Monte Carlo state estimation Hajime Fujita and Shin Ishii, Nara Institute of Science and Technology 8916 5 Takayama, Ikoma, 630 0192 JAPAN
More informationSRNDNA Model Fitting in RL Workshop
SRNDNA Model Fitting in RL Workshop yael@princeton.edu Topics: 1. trial-by-trial model fitting (morning; Yael) 2. model comparison (morning; Yael) 3. advanced topics (hierarchical fitting etc.) (afternoon;
More informationan introduction to bayesian inference
with an application to network analysis http://jakehofman.com january 13, 2010 motivation would like models that: provide predictive and explanatory power are complex enough to describe observed phenomena
More informationStatistical Learning. Philipp Koehn. 10 November 2015
Statistical Learning Philipp Koehn 10 November 2015 Outline 1 Learning agents Inductive learning Decision tree learning Measuring learning performance Bayesian learning Maximum a posteriori and maximum
More informationContext tree models for source coding
Context tree models for source coding Toward Non-parametric Information Theory Licence de droits d usage Outline Lossless Source Coding = density estimation with log-loss Source Coding and Universal Coding
More informationReinforcement Learning as Variational Inference: Two Recent Approaches
Reinforcement Learning as Variational Inference: Two Recent Approaches Rohith Kuditipudi Duke University 11 August 2017 Outline 1 Background 2 Stein Variational Policy Gradient 3 Soft Q-Learning 4 Closing
More informationPreliminary Statistics Lecture 2: Probability Theory (Outline) prelimsoas.webs.com
1 School of Oriental and African Studies September 2015 Department of Economics Preliminary Statistics Lecture 2: Probability Theory (Outline) prelimsoas.webs.com Gujarati D. Basic Econometrics, Appendix
More informationLecture 3: More on regularization. Bayesian vs maximum likelihood learning
Lecture 3: More on regularization. Bayesian vs maximum likelihood learning L2 and L1 regularization for linear estimators A Bayesian interpretation of regularization Bayesian vs maximum likelihood fitting
More informationReinforcement Learning
Reinforcement Learning Model-Based Reinforcement Learning Model-based, PAC-MDP, sample complexity, exploration/exploitation, RMAX, E3, Bayes-optimal, Bayesian RL, model learning Vien Ngo MLR, University
More information