hmm intro

The Hidden Markov Model

Markov chain
• Useful when we need to compute a probability for a
sequence of observable events
If the events hidden? then
A hidden Markov model (HMM) allows us to talk about both

observed events (like words that we see in the input) and hidden
events (like part-of-speech tags) that we think of as causal factors in
our probabilistic model.
• A hidden Markov model is a tool for representing probability
distributions over sequences of observations.
• It is a statistical model with hidden states.
• A set of states each of which has a limited number of transitions

and emissions,
• Each transition between states has an assigned probability,
• Each model starts from the start state and ends in the end state.
Hidden Markov Model
Hidden Markov Model
For Example,
Evidence –E and X is the hidden variable,
→E--X
Markov Models
Assume there are Three types of weather
Rainy
Sunny Foggy
Weather prediction is about the what be weather tomorrow,

Based on the observations on the passed
Weather at day n is,
qn - depends on known weather of the past days

(qn-1, qn-2,..)
Markov Models
We want to find that :
P(qn | qn-1, qn-2,.., q1)

This means that given the past whether what is the probability of any possible weather
of today.
For Example,
if we knew the weather for the last three days was:
{ sunny, sunny, rainy }
The probability that tomorrow would be foggy is:

P(q4= Foggy| q3 =Rainy, q2=sunny, q1=sunny )
The weather depends on today (First-order Markov Model)
Markov Model Assumption
If Weather yesterday was rainy and today is

Foggy what is the probability that tomorrow
it will be sunny
0.6
The Markov Model
0.4 0.6
0.3
Cloud
Rain y
0.2 Quest: What is the probability that the weather
for the next 7 days will be Sun-Sun-Rain- Rain-
0.1 0.2
0.3 0.1 Sun-Cloudy-Sun “ when today is sunny?
Sunny
States : S1: Rainy, S2 :cloudy , S3: Sunny
P(O|Model)=P(S3, S3, S1, S1, S3, S2, S3 |Model)
0.8
= P(S3 ).P(S3| S3) . P(S3| S3) . P(S1| S3) . P(S1| S1) . P(S3| S1) . P(S2| S3). P(S3| S2)
= π3 . α33 . α33 . α31 . α11. α13. α32. α23
= 1 (0.8)(0.8)(0.1)(0.4)(0.3)(0.1)(0.2) = 1.536 x10-4

The Markov Model
The initial State Probability matrix
0.5
π = πi = 0.2
0.3
The State transition Probability matrix

0.6 0.2 0.2
A = aij = 0.5 0.3 0.2
0.4 0.1 0.5
Quest: What is the probability of 5 consecutive
up days?
P(O|Model)=P(1,1,1,1,1|Model) = π3 . α11 . α11 . α11 . α11
= 0.5 (0.6)(0.6)(0.6)(0.6) = 0.0648
Observations
• An observation is termed as the data which is known and can be
observed.
Two HIDDEN states, ‘Rainy’ and ‘Sunny’
‘Walk’, ‘Shop’, and ‘Clean’- OBSERVATIONS

Five important components of HMM
• Initial probability distribution- The initialization distribution defines
each hidden variable in its initial condition at time t=0 (the initial
hidden state).
• One or more hidden states
• Transition probability distribution - The transition matrix is used to

show the hidden state to hidden state transition probabilities.
• A sequence of observations
• Emission probabilities - A sequence of observation likelihoods, also

called emission probabilities.
Components of HMM
An HMM is specified by the following components:
Three fundamental problems in HMM
The situation...
•Casino has two dice, one fair (F) and one loaded (L)
•Probabilities for the fair die: P(1) = P(2) = P(3) = P(4) = P(5) = P(6) = 1/6
•Probabilities for the loaded die: P(1) = P(2) = P(3) = P(4) = P(5) = 1/10; P(6) = ½
•Before each roll, the casino player switches from the fair die to the loaded die (or vice
versa) with probability 1/20
The game...
•You bet $1
•You roll (always with the fair die)
•Casino player rolls (maybe with the fair die, maybe with the loaded die)
•Player who rolls the highest number wins $2

Dishonest Casino HMM
P(1 | F) = 1/6 P(1 | L) = 1/10
P(2 | F) = 1/6 P(2 | L) = 1/10
0.95 0.05 P(3 | F) = 1/6 P(3 | L) = 1/10
0.95
P(4 | F) = 1/6 P(4 | L) = 1/10
P(5 | F) = 1/6 P(5 | L) = 1/10
FAIR LOADED P(6 | F) = 1/6 P(6 | L) = 1/2
0.05
•Evaluation: What is the probability of a sequence of outputs of an HMM?
•Decoding: Given a sequence of outputs of an HMM, what is the most probable

sequence of states that the HMM went through to produce the output?
•Learning: Given a sequence of outputs of an HMM, how do we estimate the

parameters of the model?
Evaluation Question
Suppose the casino player rolls the following sequence...
1245526462146146136136661
6646616366163661636165156
15115146123562 344
Probability = 1.3 x 10-35 (Note (1/6)67 = 7.309139054 x 10-53.)
How likely is this sequence given our model of how the casino operates?
Decoding Question
1245526462146146136136661
6646616366163661636165156
15115146123562 344
LOADED
FAIR
What portion of the sequence was generated with the fair die, and what portion
with the loaded die?
Learning Question
1245526462146146136136661
6646616366163661636165156
15115146123562 344
LOADED
P(6) = 0.64
FAIR
How “loaded” is the loaded die? How “fair” is the fair die? How often does the
casino player change from fair to loaded, and back? That is, what are the
parameters of the model?

hmm intro

Uploaded by

Copyright:

Available Formats

hmm intro

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

hmm intro

Uploaded by

Copyright:

Available Formats

The Hidden Markov Model

The Hidden Markov Model

If the events hidden? then

A hidden Markov model (HMM) allows us to talk about both

• A set of states each of which has a limited number of transitions

• Each transition between states has an assigned probability,

Hidden Markov Model

Weather prediction is about the what be weather tomorrow,

Weather at day n is,

qn - depends on known weather of the past days

P(qn | qn-1, qn-2,.., q1)

The probability that tomorrow would be foggy is:

If Weather yesterday was rainy and today is

= π3 . α33 . α33 . α31 . α11. α13. α32. α23

= 1 (0.8)(0.8)(0.1)(0.4)(0.3)(0.1)(0.2) = 1.536 x10-4

The State transition Probability matrix

Two HIDDEN states, ‘Rainy’ and ‘Sunny’

‘Walk’, ‘Shop’, and ‘Clean’- OBSERVATIONS

• One or more hidden states

• Transition probability distribution - The transition matrix is used to

• Emission probabilities - A sequence of observation likelihoods, also

•You roll (always with the fair die)

•Player who rolls the highest number wins $2

•Decoding: Given a sequence of outputs of an HMM, what is the most probable

•Learning: Given a sequence of outputs of an HMM, how do we estimate the

Probability = 1.3 x 10-35 (Note (1/6)67 = 7.309139054 x 10-53.)

You might also like

Pfad - The Proxy pFad of © 2024 Garber Painting. All rights reserved.