Upload
nhung
View
58
Download
0
Embed Size (px)
DESCRIPTION
LECTURE 10: EXPECTATION MAXIMIZATION (EM). Objectives: Jensen’s Inequality (Special Case) EM Theorem Proof EM Example – Missing Data Application: Hidden Markov Models Resources: Wiki: EM History T.D.: Brown CS Tutorial UIUC: Tutorial F.J.: Statistical Methods. - PowerPoint PPT Presentation
Citation preview
ECE 8443 – Pattern RecognitionECE 8527 – Introduction to Machine Learning and Pattern Recognition
• Objectives:Jensen’s Inequality (Special Case)EM Theorem ProofEM Example – Missing DataApplication: Hidden Markov Models
• Resources:Wiki: EM HistoryT.D.: Brown CS TutorialUIUC: TutorialF.J.: Statistical Methods
LECTURE 10: EXPECTATION MAXIMIZATION (EM)
ECE 8527: Lecture 10, Slide 2
The Expectation Maximization Algorithm (Preview)
ECE 8527: Lecture 10, Slide 3
The Expectation Maximization Algorithm (Cont.)
ECE 8527: Lecture 10, Slide 4
The Expectation Maximization Algorithm
ECE 8527: Lecture 10, Slide 5
• Expectation maximization (EM) is an approach that is used in many ways to find maximum likelihood estimates of parameters in probabilistic models.
• EM is an iterative optimization method to estimate some unknown parameters given measurement data. Used in a variety of contexts to estimate missing data or discover hidden variables.
• The intuition behind EM is an old one: alternate between estimating the unknowns and the hidden variables. This idea has been around for a long time. However, in 1977, Dempster, et al., proved convergence and explained the relationship to maximum likelihood estimation.
• EM alternates between performing an expectation (E) step, which computes an expectation of the likelihood by including the latent variables as if they were observed, and a maximization (M) step, which computes the maximum likelihood estimates of the parameters by maximizing the expected likelihood found on the E step. The parameters found on the M step are then used to begin another E step, and the process is repeated.
• This approach is the cornerstone of important algorithms such as hidden Markov modeling and discriminative training, and has been applied to fields including human language technology and image processing.
Synopsis
ECE 8527: Lecture 10, Slide 6
Lemma: If p(x) and q(x) are two discrete probability distributions, then:
with equality if and only if p(x) = q(x) for all x.
Proof:
The last step follows using a bound for the natural logarithm: .
Special Case of Jensen’s Inequality
xx
xqxpxpxp )(log)()(log)(
xx
x
x
x
xx
xpxqxp
xpxqxp
xpxqxp
xqxpxp
xqxpxp
xqxpxpxp
)1)()()((
)()(log)(
0)()(log)(
0)()(log)(
0)()(log)(
0)(log)()(log)(
1ln xx
ECE 8527: Lecture 10, Slide 7
Continuing in efforts to simplify:
We note that since both of these functions are probability distributions, they must sum to 1.0. Therefore, the inequality holds.
The general form of Jensen’s inequality relates a convex function of an integral to the integral of the convex function and is used extensively in information theory.
Special Case of Jensen’s Inequality
x x xxxx
xpxqxpxpxqxp
xpxqxp
xpxqxp ..0)()()(
)()()()1
)()()((
)()(log)(
ECE 8527: Lecture 10, Slide 8
Theorem: If then .
Proof: Let y denote observable data. Let be the probability distribution
of y under some model whose parameters are denoted by .
Let be the corresponding distribution under a different setting .
Our goal is to prove that y is more likely under than .
Let t denote some hidden, or latent, parameters that are governed by the
values of . Because is a probability distribution that sums to 1, we
can write:
Because we can exploit the dependence of y on t and using well-known
properties of a conditional probability distribution.
The EM Theorem
ytPytPytPytPtt
loglog yPyP
yP
yP
ytP
tt
yPytPyPytPyPyP loglogloglog
ECE 8527: Lecture 10, Slide 9
We can multiply each term by “1”:
where the inequality follows from our lemma.
Explanation: What exactly have we shown? If the last quantity is greater than
zero, then the new model will be better than the old model. This suggests a
strategy for finding the new parameters, θ: choose them to make the last
quantity positive!
Proof Of The EM Theorem
tt
tt
tt
tt
tt
ytPytPytPytP
ytPytPytPytP
ytPytPytPytP
ytPytPytP
ytPytPytP
ytPytPyPytP
ytPytPyPytPyPyP
,log,log
,loglog
,log,log
,log,log
,,log
,,logloglog
ECE 8527: Lecture 10, Slide 10
Discussion
• If we start with the parameter setting , and find a parameter setting for
which our inequality holds, then the observed data, y, will be more probable
under than .
• The name Expectation Maximization comes about because we take the
expectation of with respect to the old distribution and then
maximize the expectation as a function of the argument .
• Critical to the success of the algorithm is the choice of the proper
intermediate variable, t, that will allow finding the maximum of the
expectation of .
• Perhaps the most prominent use of the EM algorithm in pattern recognition is
to derive the Baum-Welch reestimation equations for a hidden Markov model.
• Many other reestimation algorithms have been derived using this approach.
ytP , ytP ,
ytPytPt
log
ECE 8527: Lecture 10, Slide 11
Example: Estimating Missing Data
2*,
22
,01,
20
,,, 4321 xxxxD
222121 Tθ
• Consider a data set with a missing element:
• Let us estimate the value of the missing point assuming a Gaussian model
with a diagonal covariance and arbitrary means:
• Expectation step:
Assuming normal distributions as initial conditions, this can be simplified to:
ytPytPt
log
41
4141
41
413
1
41420
413
14
4
44
ln(nl
)4;((ln(ln);(
dxxd
xp
xp
xpp
dxxxpppQ
kk
kk
0
0
θ
θθθx
θθxθxθθ
3
1212
2
22
21
21 )2ln(
2)4(
21(ln);(
kkpQ
θxθθ
ECE 8527: Lecture 10, Slide 12
Example: Gaussian Mixtures
• An excellent tutorial on Gaussian mixture estimation can be found at
J. Bilmes, EM Estimation
• An interactive demo showing convergence of the estimate can be found at
I. Dinov, Demonstration
ECE 8527: Lecture 10, Slide 13
Introduction To Hidden Markov Models
ECE 8527: Lecture 10, Slide 14
Introduction To Hidden Markov Models (Cont.)
ECE 8527: Lecture 10, Slide 15
Introduction To Hidden Markov Models (Cont.)
ECE 8527: Lecture 10, Slide 16
Summary• Expectation Maximization (EM) Algorithm: a generalization of Maximum
Likelihood Estimation (MLE) based on maximization of a posterior that data was generated by a model. EM is a special case of Jensen’s inequality.
• Jensen’s Inequality: describes a relationship between two probability distributions in terms of an entropy-like quantity. A key tool in proving that EM estimation converges.
• The EM Theorem: proved that estimation of a model’s parameters using an iteration of EM increases the posterior probability that the data was generated by the model.
• Demonstrated an application of the EM Theorem to the problem of estimating missing data point.
• Explained how EM can be used to reestimate parameters in a pattern recognition system.
• Introduced the concept of a hidden Markov model and explained how we will use EM to estimate the parameters of this model.