### Tags: example, examples, hidden, hmm, hmms, iam, markov, matlab, models, programming, realize, seeking, sing, whythey

# Seeking Simple HMM Example(s)

On Programmer » Matlab

4,955 words with 1 Comments; publish: Wed, 07 May 2008 09:52:00 GMT; (20029.30, « »)

I am trying to teach myself Hidden Markov Models [HMMs]. I

am sing some simple examples to understand how and why

they are used.

I realize that this is a strange request, instead of begging

for code, I'm effectively asking for data. (A good

discussion would do to.)

In a Hidden Markov Model, a measurement is OBSERVED, and

this hints at the true STATE of an underlying variable.

Supposedly, the STATE is hidden from you.

The problem is that HMMs are usually explained in terms of

flipping coins. The coins are fair, and there is nothing

"hidden" about them, so the examples are wasted on me. If I

OBSERVE that the coin is a head-up, then the STATE is head-up.

There is another common example, colored balls in urns.

Again, there seems no possibility for error in the

measurement. You look at the ball, you know the color. Why

would you need a model?

Hidden Markov Models are important because they form the

basis for many other graphical techniques for Pattern

Classification -- Bayes Nets, Factor Graphs, Markov Fields,

etc...

Here is one slightly better example. Suppose you were

trying to figure out information about a road. You want to

know if it is a small street or a large highway. So you

observe the vehicles that go by -- some are cars, some are

trucks. In this way you are observing something which is

not the actual state of the road. However, I would never

use an HMM for this.

Another example might be if you were looking through foggy

glass, or your camera was out of focus. The examples I s

must have the possibility that what you observed might be

the wrong state of affairs, and it must be applicable for

HMMs, so that I can program it up and understand it.

Thanks for any help. Best Regards...

*http://matlab.todaysummary.com/q_matlab_53663.html*

All Comments

Leave a comment...

- 1 Comments
- "Pat Finder" <pfinder.nospam.matlab.todaysummary.com.netacc.net> wrote in message
<fsb812$aep$1.matlab.todaysummary.com.fred.mathworks.com>...

> I am trying to teach myself Hidden Markov Models [HMMs].

I

> am sing some simple examples to understand how and why

> they are used.

> I realize that this is a strange request, instead of

begging

> for code, I'm effectively asking for data. (A good

> discussion would do to.)

> In a Hidden Markov Model, a measurement is OBSERVED, and

> this hints at the true STATE of an underlying variable.

> Supposedly, the STATE is hidden from you.

> The problem is that HMMs are usually explained in terms of

> flipping coins. The coins are fair, and there is nothing

> "hidden" about them, so the examples are wasted on me.

If I

> OBSERVE that the coin is a head-up, then the STATE is

head-up.

> There is another common example, colored balls in urns.

> Again, there seems no possibility for error in the

> measurement. You look at the ball, you know the color.

Why

> would you need a model?

> Hidden Markov Models are important because they form the

> basis for many other graphical techniques for Pattern

> Classification -- Bayes Nets, Factor Graphs, Markov

Fields,

> etc...

> Here is one slightly better example. Suppose you were

> trying to figure out information about a road. You want

to

> know if it is a small street or a large highway. So you

> observe the vehicles that go by -- some are cars, some are

> trucks. In this way you are observing something which is

> not the actual state of the road. However, I would never

> use an HMM for this.

> Another example might be if you were looking through foggy

> glass, or your camera was out of focus. The examples I

s

> must have the possibility that what you observed might be

> the wrong state of affairs, and it must be applicable for

> HMMs, so that I can program it up and understand it.

> Thanks for any help. Best Regards...

The analogy that I liked best was to assume you were living

in a cave, and had with you a hydrometer. You couldn't see

the outside world, but could see the readings of the

hydrometer. The underlying variable was whether or not it

was raining on the outside of the cave, what you could

observe was the moisture content of the air in the cave.

Not quite what you were after, but I hope it helps

Regards

Dave Robinson

#1; Wed, 07 May 2008 09:53:00 GMT

- "Pat Finder" <pfinder.nospam.matlab.todaysummary.com.netacc.net> wrote in message