I will try to make some clarifications below:

> While they haven't spelled it out in detail, I think
> they're using a simple logic system and definition of
> "consistent".  In other words, they're not trying to
> allow for an "agent belief" representation where we
> acknowledge that beliefs might be wrong.  Instead, if
> you assert a sentance it must be true or you've
> introduced a contradiction.

Maybe our sources are different. AFAIK a contradiction is the negation
of a tautology (P. Tidman and H. Kahane, Logic and Philosophy: A Modern
Introduction, 8th edition, p48). For example, p^~p is a contradiction.
So when I assert a contingently false sentence whose truth depends on
the world, I introduce no contradiction. But maybe our definitions of
contradiction differ.

> If we take 3 sentances:
>
> A = B.
> B = C.
> A != C.
>
> If we try to assert all three sentances, we cannot do
> it without creating a contradiction.  Since this is
> the definition of "consistency", they would say that
> we cannot "consistently" assert all three sentances,
> since doing so would introduce a contradiction.

It is true that we would introduce inconsistency, but we would not
introduce a contradiction. Of course, any set of sentences that contains
any contradictory sentence is inconsistent, but not vice versa.

> In this case, as an agent **I** can consistently
> assert the sentance "Agent A cannot assert this
> sentance without being wrong" without introducing a
> contradiction.  However if Agent A tries to assert the
> same sentance, he runs into a problem.  If he asserts
> it as true and is right, then the sentance is false,
> so he's wrong.

I know what you are trying to say, but that I don't think we are talking
about the same thing. The authors clearly intend to show two things:
1. The sentence is a tautology (i.e. necessarily true).
2. Yet, J. R. Lucas cannot assert it.
They offer two arguments to show (1). What I do not understand is the
their second argument to show (1). You are talking about (2), which I
find no problem.

> But the whole point of the illustration is simply to
> show that **sometimes** one agent is unable to
> assert/know/do things that another agent can, but that
> doesn't automatically imply that agent is inferior,
> it's inability to assert/know/do may be related to the
> specific situation.

I do understand the point; it is simply that particular argument (the
one on p950 in parentheses) that confuses me. As a philosophical
treatment, I think the chapter should reasonably be taken literally, but
when I do so, I cannot make sense of that particular argument.

Paul

Hi,
   could anyone plz let me know the link to the answers of the
exercise
of this book

In page97, line 7:
The restriction is to choose an h function that never overestimates
the cost to reach the goal. Such an h is called an admissible
heuristic. Admissible heuristics are by nature optimistic, because
they think the cost of solving the problem is less than it actually is.

Can anyone give me more explanation why it always gets the optimial
result when it never overestimates the total cost.

In page97, line 7:
The restriction is to choose an h function that never overestimates
the cost to reach the goal. Such an h is called an admissible
heuristic. Admissible heuristics are by nature optimistic, because
they think the cost of solving the problem is less than it actually is.

Can anyone give me more explanation why it always gets the optimial
result when it never overestimates the total cost.

I think my version is different from yours, but I
assume you're talking about the A* search algorithm.

The proof is in the book a page or so later, but  look
at it the other way for a second - if the path
estimate were sometimes too high, then based on the
inflated estimate you might ignore a path that would
have turned out to have a "short cut" in it.  But by
guaranteeing that the actual cost will always be more
than your estimate, you're guaranteed never to ignore
a short cut.

To put it another way, with an admissible heuristic
any unexplored path is guaranteed to be worse than or
equal to it's estimate - never better.  Thus when you
actually explore a path, you're guaranteed that it's
cost will only get worse.  So if you've found an
actual path solution that's equal to or better than
the best unexplored path estimates, the actual path
you've found is guaranteed to be the best because the
unexplored paths can only get more costly when they're
explored.

I hope that made sense.

Rob G.

--- lwudong <wudongs@...> wrote:

> In page97, line 7:
> The restriction is to choose an h function that
> never overestimates
> the cost to reach the goal. Such an h is called an
> admissible
> heuristic. Admissible heuristics are by nature
> optimistic, because
> they think the cost of solving the problem is less
> than it actually is.
>
> Can anyone give me more explanation why it always
> gets the optimial
> result when it never overestimates the total cost.
>
>
>
>
>




__________________________________
Yahoo! Mail - PC Magazine Editors' Choice 2005
http://mail.yahoo.com

I think the discussion all boils down to the
definition of "consistent" and "contradiction".  I was
kind of hoping one of the authors would chime in at
some point to save me, but...  :-)  On page 137 they
define consistency in reference to CSP problems as "an
assignment that does not violate any constraints".
That was the reason for my A,B,C example - looking at
it as if it were a CSP.

Their argument assumes the sentance must either be
"true" or "false", (no quantuum physics or "unknown"
values allowed) and hinges on the
belief/assumption/fact one cannot "consistently"
assert that a sentance is true if it can be shown to
be false.  If you buy into that, then their statement
follows naturally - if the sentance was false, then he
could not consistently assert it to be true, which
would therefore make the sentance true - a
contradiction since we assigned a value of "false" to
the sentance at the beginning.  The only way you can
assign a value to the sentance and have everything
hold together is if you assume it's true, which then
implies that everyone else can assert it without a
problem, but he cannot.

But, at this point I don't think we're going to make
any headway.  Since they didn't define the logic
system they're using, we can't really do anything
formal.

Thanks for the interesting discussion.

Rob G.


--- Paul Hsueh-Min Chang <avatar@...>
wrote:

> I will try to make some clarifications below:
>
> > While they haven't spelled it out in detail, I
> think
> > they're using a simple logic system and definition
> of
> > "consistent".  In other words, they're not trying
> to
> > allow for an "agent belief" representation where
> we
> > acknowledge that beliefs might be wrong.  Instead,
> if
> > you assert a sentance it must be true or you've
> > introduced a contradiction.
>
> Maybe our sources are different. AFAIK a
> contradiction is the negation
> of a tautology (P. Tidman and H. Kahane, Logic and
> Philosophy: A Modern
> Introduction, 8th edition, p48). For example, p^~p
> is a contradiction.
> So when I assert a contingently false sentence whose
> truth depends on
> the world, I introduce no contradiction. But maybe
> our definitions of
> contradiction differ.
>
> > If we take 3 sentances:
> >
> > A = B.
> > B = C.
> > A != C.
> >
> > If we try to assert all three sentances, we cannot
> do
> > it without creating a contradiction.  Since this
> is
> > the definition of "consistency", they would say
> that
> > we cannot "consistently" assert all three
> sentances,
> > since doing so would introduce a contradiction.
>
> It is true that we would introduce inconsistency,
> but we would not
> introduce a contradiction. Of course, any set of
> sentences that contains
> any contradictory sentence is inconsistent, but not
> vice versa.
>
> > In this case, as an agent **I** can consistently
> > assert the sentance "Agent A cannot assert this
> > sentance without being wrong" without introducing
> a
> > contradiction.  However if Agent A tries to assert
> the
> > same sentance, he runs into a problem.  If he
> asserts
> > it as true and is right, then the sentance is
> false,
> > so he's wrong.
>
> I know what you are trying to say, but that I don't
> think we are talking
> about the same thing. The authors clearly intend to
> show two things:
> 1. The sentence is a tautology (i.e. necessarily
> true).
> 2. Yet, J. R. Lucas cannot assert it.
> They offer two arguments to show (1). What I do not
> understand is the
> their second argument to show (1). You are talking
> about (2), which I
> find no problem.
>
> > But the whole point of the illustration is simply
> to
> > show that **sometimes** one agent is unable to
> > assert/know/do things that another agent can, but
> that
> > doesn't automatically imply that agent is
> inferior,
> > it's inability to assert/know/do may be related to
> the
> > specific situation.
>
> I do understand the point; it is simply that
> particular argument (the
> one on p950 in parentheses) that confuses me. As a
> philosophical
> treatment, I think the chapter should reasonably be
> taken literally, but
> when I do so, I cannot make sense of that particular
> argument.
>
> Paul
>
>
>





__________________________________
Yahoo! Mail - PC Magazine Editors' Choice 2005
http://mail.yahoo.com

---

From: aima-talk@yahoogroups.com [mailto:aima-talk@yahoogroups.com] On Behalf Of The Geek
Sent: Monday, September 26, 2005 6:03 PM
To: aima-talk@yahoogroups.com
Subject: Re: [aima-talk] about A* search

I think my version is different from yours, but I
assume you're talking about the A* search algorithm.

The proof is in the book a page or so later, but  look
at it the other way for a second - if the path
estimate were sometimes too high, then based on the
inflated estimate you might ignore a path that would
have turned out to have a "short cut" in it.  But by
guaranteeing that the actual cost will always be more
than your estimate, you're guaranteed never to ignore
a short cut. 

To put it another way, with an admissible heuristic
any unexplored path is guaranteed to be worse than or
equal to it's estimate - never better.  Thus when you
actually explore a path, you're guaranteed that it's
cost will only get worse.  So if you've found an
actual path solution that's equal to or better than
the best unexplored path estimates, the actual path
you've found is guaranteed to be the best because the
unexplored paths can only get more costly when they're
explored.

I hope that made sense.

Rob G.

--- lwudong <wudongs@...> wrote:

> In page97, line 7:
> The restriction is to choose an h function that
> never overestimates
> the cost to reach the goal. Such an h is called an
> admissible
> heuristic. Admissible heuristics are by nature
> optimistic, because
> they think the cost of solving the problem is less
> than it actually is.
>
> Can anyone give me more explanation why it always
> gets the optimial
> result when it never overestimates the total cost.
>
>
>
>
>



           
__________________________________
Yahoo! Mail - PC Magazine Editors' Choice 2005
http://mail.yahoo.com

---

From: savastinuk [mailto:minnie@...]
Sent: Tuesday, September 27, 2005 7:04 AM
To: 'aima-talk@yahoogroups.com'
Subject: RE: [aima-talk] about A* search

This makes sense. : )

 

Can you also explain consistent? Or, better yet, INconsistent?

Still talking A*.

 

thanks....

Susan

---

From: aima-talk@yahoogroups.com [mailto:aima-talk@yahoogroups.com] On Behalf Of The Geek
Sent: Monday, September 26, 2005 6:03 PM
To: aima-talk@yahoogroups.com
Subject: Re: [aima-talk] about A* search

I think my version is different from yours, but I
assume you're talking about the A* search algorithm.

The proof is in the book a page or so later, but  look
at it the other way for a second - if the path
estimate were sometimes too high, then based on the
inflated estimate you might ignore a path that would
have turned out to have a "short cut" in it.  But by
guaranteeing that the actual cost will always be more
than your estimate, you're guaranteed never to ignore
a short cut. 

To put it another way, with an admissible heuristic
any unexplored path is guaranteed to be worse than or
equal to it's estimate - never better.  Thus when you
actually explore a path, you're guaranteed that it's
cost will only get worse.  So if you've found an
actual path solution that's equal to or better than
the best unexplored path estimates, the actual path
you've found is guaranteed to be the best because the
unexplored paths can only get more costly when they're
explored.

I hope that made sense.

Rob G.

--- lwudong <wudongs@...> wrote:

> In page97, line 7:
> The restriction is to choose an h function that
> never overestimates
> the cost to reach the goal. Such an h is called an
> admissible
> heuristic. Admissible heuristics are by nature
> optimistic, because
> they think the cost of solving the problem is less
> than it actually is.
>
> Can anyone give me more explanation why it always
> gets the optimial
> result when it never overestimates the total cost.
>
>
>
>
>



           
__________________________________
Yahoo! Mail - PC Magazine Editors' Choice 2005
http://mail.yahoo.com

I think my version is different from yours, but I
assume you're talking about the A* search algorithm.

The proof is in the book a page or so later, but  look
at it the other way for a second - if the path
estimate were sometimes too high, then based on the
inflated estimate you might ignore a path that would
have turned out to have a "short cut" in it.  But by
guaranteeing that the actual cost will always be more
than your estimate, you're guaranteed never to ignore
a short cut. 

To put it another way, with an admissible heuristic
any unexplored path is guaranteed to be worse than or
equal to it's estimate - never better.  Thus when you
actually explore a path, you're guaranteed that it's
cost will only get worse.  So if you've found an
actual path solution that's equal to or better than
the best unexplored path estimates, the actual path
you've found is guaranteed to be the best because the
unexplored paths can only get more costly when they're
explored.

I hope that made sense.

Rob G.

--- lwudong <wudongs@...> wrote:

> In page97, line 7:
> The restriction is to choose an h function that
> never overestimates
> the cost to reach the goal. Such an h is called an
> admissible
> heuristic. Admissible heuristics are by nature
> optimistic, because
> they think the cost of solving the problem is less
> than it actually is.
>
> Can anyone give me more explanation why it always
> gets the optimial
> result when it never overestimates the total cost.
>
>
>
>
>



           
__________________________________
Yahoo! Mail - PC Magazine Editors' Choice 2005
http://mail.yahoo.com

__________________________________________________
Do You Yahoo!?
Tired of spam? Yahoo! Mail has the best spam protection around
http://mail.yahoo.com

Consistency of heuristics is a little more tricky to
explain, since every consistent one is also
admissible.

If you're not from the U.S., I'll apologize in advance
for the following example....

Lets say you're trying to get from New York to L.A. by
car - forget the fact that it would now cost you a
small fortune to do so.  A consistent heuristic is one
where the estimate to get from New York to L.A. must
be equal to or smaller than the actual cost to get
from New York to any other city **plus** the estimate
to get from that city to L.A.  In other words, if you
drive from New York to Chicago, then estimate the
distance from Chicago to L.A. you're not supposed to
get a smaller answer than your original estimate.  If
you use straight-line distance, it's easy to see this
is consistent.

Admissible heurstics must guarantee the estimate is no
larger than the actual cost turns out to be.
Consistent heuristics must also guarantee the "revised
estimate" (the sum of the actual distance traveled so
far plus the estimate of what you've got remaining)
never goes down as you explore the path.

You have to get kind of goofy to find things that are
admissible but not consistent - taking the
straight-line distance divided by the number of
letters in the city name for example.  The estimate is
guaranteed to be low (since it's always less than the
straight-line distance), and thus is admissible.  When
you start at New York your estimate would be 2400/7 =
342.86.  If you drove 95 miles to Philadelphia, you're
estimate from Philadelphia to L.A. would be 2320 / 12
= 193.33.  Adding that back to the 95 miles you drove
from New York we see that we now think we can get from
New York to L.A. by way of Philadelphia for an
estimated cost of 193.33 + 95 = 288.33, less than our
original estimate of 342.86, thus demonstrating that
the heuristic is not consistent.

Rob G.

--- savastinuk <minnie@...> wrote:

> This makes sense. : )
>
> Can you also explain consistent? Or, better yet,
> INconsistent?
> Still talking A*.
>
> thanks....
> Susan
>
>
>   _____
>
> From: aima-talk@yahoogroups.com
> [mailto:aima-talk@yahoogroups.com] On Behalf
> Of The Geek
> Sent: Monday, September 26, 2005 6:03 PM
> To: aima-talk@yahoogroups.com
> Subject: Re: [aima-talk] about A* search
>
>
> I think my version is different from yours, but I
> assume you're talking about the A* search algorithm.
>
> The proof is in the book a page or so later, but
> look
> at it the other way for a second - if the path
> estimate were sometimes too high, then based on the
> inflated estimate you might ignore a path that would
> have turned out to have a "short cut" in it.  But by
> guaranteeing that the actual cost will always be
> more
> than your estimate, you're guaranteed never to
> ignore
> a short cut.
>
> To put it another way, with an admissible heuristic
> any unexplored path is guaranteed to be worse than
> or
> equal to it's estimate - never better.  Thus when
> you
> actually explore a path, you're guaranteed that it's
> cost will only get worse.  So if you've found an
> actual path solution that's equal to or better than
> the best unexplored path estimates, the actual path
> you've found is guaranteed to be the best because
> the
> unexplored paths can only get more costly when
> they're
> explored.
>
> I hope that made sense.
>
> Rob G.
>
> --- lwudong <wudongs@...> wrote:
>
> > In page97, line 7:
> > The restriction is to choose an h function that
> > never overestimates
> > the cost to reach the goal. Such an h is called an
> > admissible
> > heuristic. Admissible heuristics are by nature
> > optimistic, because
> > they think the cost of solving the problem is less
> > than it actually is.
> >
> > Can anyone give me more explanation why it always
> > gets the optimial
> > result when it never overestimates the total cost.
> >
> >
> >
> >
> >
>
>
>
>
> __________________________________
> Yahoo! Mail - PC Magazine Editors' Choice 2005
> http://mail.yahoo.com
>
>
>
>
>
> SPONSORED LINKS
> Artificial
>
<http://groups.yahoo.com/gads?t=ms&k=Artificial+intelligence+software&w1=Art
>
ificial+intelligence+software&w2=Artificial+intelligence+in+business&w3=Arti
>
ficial+intelligence&w4=Artificial+intelligence+introduction&c=4&s=150&.sig=Z
> -KgstNm21fXWEV4ASPvHQ> intelligence software
> Artificial
>
<http://groups.yahoo.com/gads?t=ms&k=Artificial+intelligence+in+business&w1=
>
Artificial+intelligence+software&w2=Artificial+intelligence+in+business&w3=A
>
rtificial+intelligence&w4=Artificial+intelligence+introduction&c=4&s=150&.si
> g=477uHvbluHUGZv4M3Wsnag> intelligence in business
> Artificial
>
<http://groups.yahoo.com/gads?t=ms&k=Artificial+intelligence&w1=Artificial+i
>
ntelligence+software&w2=Artificial+intelligence+in+business&w3=Artificial+in
>
telligence&w4=Artificial+intelligence+introduction&c=4&s=150&.sig=LONI6S4JBL
> JohI0gb7t-Ug> intelligence
> Artificial
>
<http://groups.yahoo.com/gads?t=ms&k=Artificial+intelligence+introduction&w1
>
=Artificial+intelligence+software&w2=Artificial+intelligence+in+business&w3=
>
Artificial+intelligence&w4=Artificial+intelligence+introduction&c=4&s=150&.s
> ig=n5dsIRz6BWbukAv5Ru4LcA> intelligence introduction
>
>
>   _____
>
> YAHOO! GROUPS LINKS
>
>
>
> *  Visit your group "aima-talk
> <http://groups.yahoo.com/group/aima-talk> " on the
> web.
>
>
> *  To unsubscribe from this group, send an email to:
>  aima-talk-unsubscribe@yahoogroups.com
>
<mailto:aima-talk-unsubscribe@yahoogroups.com?subject=Unsubscribe>
>
>
>
> *  Your use of Yahoo! Groups is subject to the
> Yahoo! Terms of Service
> <http://docs.yahoo.com/info/terms/> .
>
>
>   _____
>
>
>




__________________________________
Yahoo! Mail - PC Magazine Editors' Choice 2005
http://mail.yahoo.com

---

From: aima-talk@yahoogroups.com [mailto:aima-talk@yahoogroups.com] On Behalf Of The Geek
Sent: Tuesday, September 27, 2005 1:46 PM
To: aima-talk@yahoogroups.com
Subject: RE: [aima-talk] about A* search

Consistency of heuristics is a little more tricky to
explain, since every consistent one is also
admissible.

If you're not from the U.S., I'll apologize in advance
for the following example....

Lets say you're trying to get from New York to L.A. by
car - forget the fact that it would now cost you a
small fortune to do so.  A consistent heuristic is one
where the estimate to get from New York to L.A. must
be equal to or smaller than the actual cost to get
from New York to any other city **plus** the estimate
to get from that city to L.A.  In other words, if you
drive from New York to Chicago, then estimate the
distance from Chicago to L.A. you're not supposed to
get a smaller answer than your original estimate.  If
you use straight-line distance, it's easy to see this
is consistent.

Admissible heurstics must guarantee the estimate is no
larger than the actual cost turns out to be.
Consistent heuristics must also guarantee the "revised
estimate" (the sum of the actual distance traveled so
far plus the estimate of what you've got remaining)
never goes down as you explore the path.

You have to get kind of goofy to find things that are
admissible but not consistent - taking the
straight-line distance divided by the number of
letters in the city name for example.  The estimate is
guaranteed to be low (since it's always less than the
straight-line distance), and thus is admissible.  When
you start at New York your estimate would be 2400/7 =
342.86.  If you drove 95 miles to Philadelphia, you're
estimate from Philadelphia to L.A. would be 2320 / 12
= 193.33.  Adding that back to the 95 miles you drove
from New York we see that we now think we can get from
New York to L.A. by way of Philadelphia for an
estimated cost of 193.33 + 95 = 288.33, less than our
original estimate of 342.86, thus demonstrating that
the heuristic is not consistent.

Rob G.

--- savastinuk <minnie@...> wrote:

> This makes sense. : )
> 
> Can you also explain consistent? Or, better yet,
> INconsistent?
> Still talking A*.
> 
> thanks....
> Susan
>
>
>   _____ 
>
> From: aima-talk@yahoogroups.com
> [mailto:aima-talk@yahoogroups.com] On Behalf
> Of The Geek
> Sent: Monday, September 26, 2005 6:03 PM
> To: aima-talk@yahoogroups.com
> Subject: Re: [aima-talk] about A* search
>
>
> I think my version is different from yours, but I
> assume you're talking about the A* search algorithm.
>
> The proof is in the book a page or so later, but
> look
> at it the other way for a second - if the path
> estimate were sometimes too high, then based on the
> inflated estimate you might ignore a path that would
> have turned out to have a "short cut" in it.  But by
> guaranteeing that the actual cost will always be
> more
> than your estimate, you're guaranteed never to
> ignore
> a short cut. 
>
> To put it another way, with an admissible heuristic
> any unexplored path is guaranteed to be worse than
> or
> equal to it's estimate - never better.  Thus when
> you
> actually explore a path, you're guaranteed that it's
> cost will only get worse.  So if you've found an
> actual path solution that's equal to or better than
> the best unexplored path estimates, the actual path
> you've found is guaranteed to be the best because
> the
> unexplored paths can only get more costly when
> they're
> explored.
>
> I hope that made sense.
>
> Rob G.
>
> --- lwudong <wudongs@...> wrote:
>
> > In page97, line 7:
> > The restriction is to choose an h function that
> > never overestimates
> > the cost to reach the goal. Such an h is called an
> > admissible
> > heuristic. Admissible heuristics are by nature
> > optimistic, because
> > they think the cost of solving the problem is less
> > than it actually is.
> >
> > Can anyone give me more explanation why it always
> > gets the optimial
> > result when it never overestimates the total cost.
> >
> >
> >
> >
> >
>
>
>
>            
> __________________________________
> Yahoo! Mail - PC Magazine Editors' Choice 2005
> http://mail.yahoo.com
>
>
>
>
>
> SPONSORED LINKS
> Artificial
>
<http://groups.yahoo.com/gads?t=ms&k=Artificial+intelligence+software&w1=Art
>
ificial+intelligence+software&w2=Artificial+intelligence+in+business&w3=Arti
>
ficial+intelligence&w4=Artificial+intelligence+introduction&c=4&s=150&.sig=Z
> -KgstNm21fXWEV4ASPvHQ> intelligence software
> Artificial
>
<http://groups.yahoo.com/gads?t=ms&k=Artificial+intelligence+in+business&w1=
>
Artificial+intelligence+software&w2=Artificial+intelligence+in+business&w3=A
>
rtificial+intelligence&w4=Artificial+intelligence+introduction&c=4&s=150&.si
> g=477uHvbluHUGZv4M3Wsnag> intelligence in business
> Artificial
>
<http://groups.yahoo.com/gads?t=ms&k=Artificial+intelligence&w1=Artificial+i
>
ntelligence+software&w2=Artificial+intelligence+in+business&w3=Artificial+in
>
telligence&w4=Artificial+intelligence+introduction&c=4&s=150&.sig=LONI6S4JBL
> JohI0gb7t-Ug> intelligence      
> Artificial
>
<http://groups.yahoo.com/gads?t=ms&k=Artificial+intelligence+introduction&w1
>
=Artificial+intelligence+software&w2=Artificial+intelligence+in+business&w3=
>
Artificial+intelligence&w4=Artificial+intelligence+introduction&c=4&s=150&.s
> ig=n5dsIRz6BWbukAv5Ru4LcA> intelligence introduction
>      
>
>   _____ 
>
> YAHOO! GROUPS LINKS
>
>
>      
> *      Visit your group "aima-talk
> <http://groups.yahoo.com/group/aima-talk> " on the
> web.
>  
>
> *      To unsubscribe from this group, send an email to:
>  aima-talk-unsubscribe@yahoogroups.com
>
<mailto:aima-talk-unsubscribe@yahoogroups.com?subject=Unsubscribe>
>
>  
>
> *      Your use of Yahoo! Groups is subject to the
> Yahoo! Terms of Service
> <http://docs.yahoo.com/info/terms/> .
>
>
>   _____ 
>
>
>



           
__________________________________
Yahoo! Mail - PC Magazine Editors' Choice 2005
http://mail.yahoo.com

hey everybody am doing a project on image processing n am using the convolution theorem given in chapter 25.in tht am not able to get an image function.

 it would be a great help for me if anyone could tell wat sort of a function is this image function.

          thanx.

The index in my edition (2nd) gives only two mentions
of convolution: pg 869 and pg 899. These are in chapter
24, not ch. 25. So I'm not sure what you're looking at.
But in general, when you apply a convolution to an image,
the "image function" is just the pixel value at each
(x,y) location. The value you use will depend on context
- what you want to do. A commonly used value is pixel
brightness. A white pixel has brightness 255, and a black
one has brightness 0. If you're starting from a color
image, you can calculate brightness by averaging the red,
blue, and green values at pixel (x,y).

Here's a good explanation of convolution in an image-
processing context:
http://www.cee.hw.ac.uk/hipr/html/convolve.html

- Robin


--- In aima-talk@yahoogroups.com, shivam shukla <shirohin@y...> wrote:
>
> hey everybody am doing a project on image processing n am using the
convolution theorem given in chapter 25.in tht am not able to get an
image function.
>
>  it would be a great help for me if anyone could tell wat sort of a
function is this image function.
>
>           thanx.
>
>
> ---------------------------------
>  Yahoo! India Matrimony: Find your partner now.

I doubt if I can make the issues more clear, but I'll try.

I still think the main problem lies in the phrase "if it were false". If
a proposition in a set is "shown" to be false from observation rather
than through logical proof, the proposition does not make the set
inconsistent, because it *could* be true in some other possible world.
If I recall correctly, that is how logical possibility is defined. Your
ABC example is indeed inconsistent because it can be proved without
observing the world state that the set contains at least one false sentence.

So, according to the defintion of consistency, the authors mostly likely
do not mean "if it were provably false". Otherwise, the argument would
be employing a tautologus sentence: "if the sentence were provably
false, then he (or in fact, anyone) could not consistently assert it to
be true". One derives nothing from a tautology

But, without the word "provably", I don't see how the contingent truth
or falsity of a sentence have anything to do with consistency, for
reasons above. Furthermore, it seems to me that J. R. Lucas just cannot
consistently assert the sentence anyway, whether it is true or false.

I do agree that we can do nothing formal without identifying the
underlying logic system. Perhaps somebody can enlighten us?

  > Thanks for the interesting discussion.
Thank you.

Paul



The Geek wrote:

> I think the discussion all boils down to the
> definition of "consistent" and "contradiction".  I was
> kind of hoping one of the authors would chime in at
> some point to save me, but...  :-)  On page 137 they
> define consistency in reference to CSP problems as "an
> assignment that does not violate any constraints".
> That was the reason for my A,B,C example - looking at
> it as if it were a CSP.
>
> Their argument assumes the sentance must either be
> "true" or "false", (no quantuum physics or "unknown"
> values allowed) and hinges on the
> belief/assumption/fact one cannot "consistently"
> assert that a sentance is true if it can be shown to
> be false.  If you buy into that, then their statement
> follows naturally - if the sentance was false, then he
> could not consistently assert it to be true, which
> would therefore make the sentance true - a
> contradiction since we assigned a value of "false" to
> the sentance at the beginning.  The only way you can
> assign a value to the sentance and have everything
> hold together is if you assume it's true, which then
> implies that everyone else can assert it without a
> problem, but he cannot.
>
> But, at this point I don't think we're going to make
> any headway.  Since they didn't define the logic
> system they're using, we can't really do anything
> formal.
>
> Thanks for the interesting discussion.
>
> Rob G.
>
>
> --- Paul Hsueh-Min Chang <avatar@...>
> wrote:
>
> > I will try to make some clarifications below:
> >
> > > While they haven't spelled it out in detail, I
> > think
> > > they're using a simple logic system and definition
> > of
> > > "consistent".  In other words, they're not trying
> > to
> > > allow for an "agent belief" representation where
> > we
> > > acknowledge that beliefs might be wrong.  Instead,
> > if
> > > you assert a sentance it must be true or you've
> > > introduced a contradiction.
> >
> > Maybe our sources are different. AFAIK a
> > contradiction is the negation
> > of a tautology (P. Tidman and H. Kahane, Logic and
> > Philosophy: A Modern
> > Introduction, 8th edition, p48). For example, p^~p
> > is a contradiction.
> > So when I assert a contingently false sentence whose
> > truth depends on
> > the world, I introduce no contradiction. But maybe
> > our definitions of
> > contradiction differ.
> >
> > > If we take 3 sentances:
> > >
> > > A = B.
> > > B = C.
> > > A != C.
> > >
> > > If we try to assert all three sentances, we cannot
> > do
> > > it without creating a contradiction.  Since this
> > is
> > > the definition of "consistency", they would say
> > that
> > > we cannot "consistently" assert all three
> > sentances,
> > > since doing so would introduce a contradiction.
> >
> > It is true that we would introduce inconsistency,
> > but we would not
> > introduce a contradiction. Of course, any set of
> > sentences that contains
> > any contradictory sentence is inconsistent, but not
> > vice versa.
> >
> > > In this case, as an agent **I** can consistently
> > > assert the sentance "Agent A cannot assert this
> > > sentance without being wrong" without introducing
> > a
> > > contradiction.  However if Agent A tries to assert
> > the
> > > same sentance, he runs into a problem.  If he
> > asserts
> > > it as true and is right, then the sentance is
> > false,
> > > so he's wrong.
> >
> > I know what you are trying to say, but that I don't
> > think we are talking
> > about the same thing. The authors clearly intend to
> > show two things:
> > 1. The sentence is a tautology (i.e. necessarily
> > true).
> > 2. Yet, J. R. Lucas cannot assert it.
> > They offer two arguments to show (1). What I do not
> > understand is the
> > their second argument to show (1). You are talking
> > about (2), which I
> > find no problem.
> >
> > > But the whole point of the illustration is simply
> > to
> > > show that **sometimes** one agent is unable to
> > > assert/know/do things that another agent can, but
> > that
> > > doesn't automatically imply that agent is
> > inferior,
> > > it's inability to assert/know/do may be related to
> > the
> > > specific situation.
> >
> > I do understand the point; it is simply that
> > particular argument (the
> > one on p950 in parentheses) that confuses me. As a
> > philosophical
> > treatment, I think the chapter should reasonably be
> > taken literally, but
> > when I do so, I cannot make sense of that particular
> > argument.
> >
> > Paul

anyone can give me some clues about the up-to-date info about expert
system? Seems this topic has been being faded?

Thanks in advance.

"oldatum" <oldatum@...> writes:

> anyone can give me some clues about the up-to-date info about expert
> system? Seems this topic has been being faded?

I wondered this myself, and posted some notes to my blog:

   Expert systems: what happened in the past decade?
   http://www.paoloamoroso.it/log/050309.html


Paolo
--
Lisp Propulsion Laboratory log - http://www.paoloamoroso.it/log

Hello ML,
I'm the guy who's doing an alternative implementation of the algorithms in the
book in Java. I've done a long break but now I've decided to continue with the
project, even with no help :-(

I've just finished the package related to chapter 7.

You can see the documentation in:
http://www.artificialidea.com/my_programs/docs/index.html

If you would like to download all I have done until now:
http://www.artificialidea.com/index.php?page=my_programs

I'd more than appreciate any comment.

Regards,
--
Ivan F. Villanueva B.
The dream of intelligent machines: www.artificialidea.com
Encrypted mail preferred.
GPG Key Id: 3FDBF85F 2004-10-18 Ivan-Fernando Villanueva Barrio

Dear everyone:

Could anyone please tell me if I would like to do research on Automatic Music Composition Generator, who is the professor in this area and what is the website to focus on?

Thank you for all answers.

Best regards,

Jaidee, Mr.

Yahoo! Music Unlimited - Access over 1 million songs. Try it free.

YAHOO! GROUPS LINKS

•  Visit your group "aima-talk" on the web.
   
•  To unsubscribe from this group, send an email to:
   aima-talk-unsubscribe@yahoogroups.com
   
•  Your use of Yahoo! Groups is subject to the Yahoo! Terms of Service.

Rob,

 

Thanks so much! This helps me with a homework problem that I was completely stumped on. I'll cite your letter, as our teacher asked us to do if we get help with an answer.

 

Your admissible but inconsistent trip example went right past where I live, near Philadelphia. : )

 

regards,

Susan

---

From: aima-talk@yahoogroups.com [mailto:aima-talk@yahoogroups.com] On Behalf Of The Geek
Sent: Tuesday, September 27, 2005 1:46 PM
To: aima-talk@yahoogroups.com
Subject: RE: [aima-talk] about A* search

Consistency of heuristics is a little more tricky to
explain, since every consistent one is also
admissible.

If you're not from the U.S., I'll apologize in advance
for the following example....

Lets say you're trying to get from New York to L.A. by
car - forget the fact that it would now cost you a
small fortune to do so.  A consistent heuristic is one
where the estimate to get from New York to L.A. must
be equal to or smaller than the actual cost to get
from New York to any other city **plus** the estimate
to get from that city to L.A.  In other words, if you
drive from New York to Chicago, then estimate the
distance from Chicago to L.A. you're not supposed to
get a smaller answer than your original estimate.  If
you use straight-line distance, it's easy to see this
is consistent.

Admissible heurstics must guarantee the estimate is no
larger than the actual cost turns out to be.
Consistent heuristics must also guarantee the "revised
estimate" (the sum of the actual distance traveled so
far plus the estimate of what you've got remaining)
never goes down as you explore the path.

You have to get kind of goofy to find things that are
admissible but not consistent - taking the
straight-line distance divided by the number of
letters in the city name for example.  The estimate is
guaranteed to be low (since it's always less than the
straight-line distance), and thus is admissible.  When
you start at New York your estimate would be 2400/7 =
342.86.  If you drove 95 miles to Philadelphia, you're
estimate from Philadelphia to L.A. would be 2320 / 12
= 193.33.  Adding that back to the 95 miles you drove
from New York we see that we now think we can get from
New York to L.A. by way of Philadelphia for an
estimated cost of 193.33 + 95 = 288.33, less than our
original estimate of 342.86, thus demonstrating that
the heuristic is not consistent.

Rob G.

--- savastinuk <minnie@...> wrote:

> This makes sense. : )
> 
> Can you also explain consistent? Or, better yet,
> INconsistent?
> Still talking A*.
> 
> thanks....
> Susan
>
>
>   _____ 
>
> From: aima-talk@yahoogroups.com
> [mailto:aima-talk@yahoogroups.com] On Behalf
> Of The Geek
> Sent: Monday, September 26, 2005 6:03 PM
> To: aima-talk@yahoogroups.com
> Subject: Re: [aima-talk] about A* search
>
>
> I think my version is different from yours, but I
> assume you're talking about the A* search algorithm.
>
> The proof is in the book a page or so later, but
> look
> at it the other way for a second - if the path
> estimate were sometimes too high, then based on the
> inflated estimate you might ignore a path that would
> have turned out to have a "short cut" in it.  But by
> guaranteeing that the actual cost will always be
> more
> than your estimate, you're guaranteed never to
> ignore
> a short cut. 
>
> To put it another way, with an admissible heuristic
> any unexplored path is guaranteed to be worse than
> or
> equal to it's estimate - never better.  Thus when
> you
> actually explore a path, you're guaranteed that it's
> cost will only get worse.  So if you've found an
> actual path solution that's equal to or better than
> the best unexplored path estimates, the actual path
> you've found is guaranteed to be the best because
> the
> unexplored paths can only get more costly when
> they're
> explored.
>
> I hope that made sense.
>
> Rob G.
>
> --- lwudong <wudongs@...> wrote:
>
> > In page97, line 7:
> > The restriction is to choose an h function that
> > never overestimates
> > the cost to reach the goal. Such an h is called an
> > admissible
> > heuristic. Admissible heuristics are by nature
> > optimistic, because
> > they think the cost of solving the problem is less
> > than it actually is.
> >
> > Can anyone give me more explanation why it always
> > gets the optimial
> > result when it never overestimates the total cost.
> >
> >
> >
> >
> >
>
>
>
>            
> __________________________________
> Yahoo! Mail - PC Magazine Editors' Choice 2005
> http://mail.yahoo.com
>
>
>
>
>
> SPONSORED LINKS
> Artificial
>
<http://groups.yahoo.com/gads?t=ms&k=Artificial+intelligence+software&w1=Art
>
ificial+intelligence+software&w2=Artificial+intelligence+in+business&w3=Arti
>
ficial+intelligence&w4=Artificial+intelligence+introduction&c=4&s=150&.sig=Z
> -KgstNm21fXWEV4ASPvHQ> intelligence software
> Artificial
>
<http://groups.yahoo.com/gads?t=ms&k=Artificial+intelligence+in+business&w1=
>
Artificial+intelligence+software&w2=Artificial+intelligence+in+business&w3=A
>
rtificial+intelligence&w4=Artificial+intelligence+introduction&c=4&s=150&.si
> g=477uHvbluHUGZv4M3Wsnag> intelligence in business
> Artificial
>
<http://groups.yahoo.com/gads?t=ms&k=Artificial+intelligence&w1=Artificial+i
>
ntelligence+software&w2=Artificial+intelligence+in+business&w3=Artificial+in
>
telligence&w4=Artificial+intelligence+introduction&c=4&s=150&.sig=LONI6S4JBL
> JohI0gb7t-Ug> intelligence      
> Artificial
>
<http://groups.yahoo.com/gads?t=ms&k=Artificial+intelligence+introduction&w1
>
=Artificial+intelligence+software&w2=Artificial+intelligence+in+business&w3=
>
Artificial+intelligence&w4=Artificial+intelligence+introduction&c=4&s=150&.s
> ig=n5dsIRz6BWbukAv5Ru4LcA> intelligence introduction
>      
>
>   _____ 
>
> YAHOO! GROUPS LINKS
>
>
>      
> *      Visit your group "aima-talk
> <http://groups.yahoo.com/group/aima-talk> " on the
> web.
>  
>
> *      To unsubscribe from this group, send an email to:
>  aima-talk-unsubscribe@yahoogroups.com
>
<mailto:aima-talk-unsubscribe@yahoogroups.com?subject=Unsubscribe>
>
>  
>
> *      Your use of Yahoo! Groups is subject to the
> Yahoo! Terms of Service
> <http://docs.yahoo.com/info/terms/> .
>
>
>   _____ 
>
>
>



           
__________________________________
Yahoo! Mail - PC Magazine Editors' Choice 2005
http://mail.yahoo.com

How can we know that our solutions are correct or not ?

hi there!
  is there any solution to the excercises of aima? like
  chapter by chapter? If there is any on the net, plz let
  it be known, and if anyone has solved any hirself, plz
  let us share.

On Tue, Oct 11, 2005 at 01:44:06AM -0700, [3!|_/|_ wrote:
> Does anyone have the implementation of A* search
> Of Romania Map or some other
> in prolog or any other

If by "any other" you mean any other language, yes there are A* implementations
in Lisp, Python and Java on the Aima webpage, and on my homepage in java at:
www.artificialidea.com/index.php?page=my_programs

Regards,
Iván.
--
Ivan F. Villanueva B.
The dream of intelligent machines: www.artificialidea.com
Encrypted mail preferred.
GPG Key Id: 3FDBF85F 2004-10-18 Ivan-Fernando Villanueva Barrio

Page 126 of AIMA, 2nd ed, notes that:

"It is easy to prove that a random walk will eventually find a goal or
complete its exploration, provided that the space is finite."

Here, a footnote continues:

"The infinite case is much more tricky. Random walks are complete on
infinite one-dimensional and two dimensional grids, but not on three
dimensional grids! In the latter case, the probability that the walk
ever returns to the starting point is only about 0.3405."

I was surprised by the claim above (random walks are complete for 1 and
2 dimensions but not for 3). Can anyone explain why this is true?

-Neil

It all began with Polya's 1921 proof:

http://mathworld.wolfram.com/PolyasRandomWalkConstants.html

Then follow the link from there "[Pages Linking Here]"

which has links to discussions of 1-, 2-, and 3-dimensional walks.

   - Bob

_______________________________________________________________
Robert P. Futrelle      | Biological Knowledge Laboratory
     Associate Professor  | College of Computer  and Information
                          |     Science MS WVH202
Office: (617)-373-4239  | Northeastern University
Fax:    (617)-373-5121  | 360 Huntington Ave.
futrelle@...    | Boston, MA 02115
           http://www.ccs.neu.edu/home/futrelle
    http://www.bionlp.org   http://www.diagrams.org

            http://biologicalknowledge.com
         mailto:biologicalknowledge@...
_______________________________________________________________


>Page 126 of AIMA, 2nd ed, notes that:
>
>"It is easy to prove that a random walk will eventually find a goal or
>complete its exploration, provided that the space is finite."
>
>Here, a footnote continues:
>
>"The infinite case is much more tricky. Random walks are complete on
>infinite one-dimensional and two dimensional grids, but not on three
>dimensional grids! In the latter case, the probability that the walk
>ever returns to the starting point is only about 0.3405."
>
>I was surprised by the claim above (random walks are complete for 1 and
>2 dimensions but not for 3). Can anyone explain why this is true?
>
>-Neil
>
>
>
>
>
>YAHOO! GROUPS LINKS
>
>  Visit your group
>"<http://groups.yahoo.com/group/aima-talk>aima-talk" on the web.
>
>  To unsubscribe from this group, send an email to:
>
><mailto:aima-talk-unsubscribe@yahoogroups.com?subject=Unsubscribe>aima-talk-uns\
ubscribe@yahoogroups.com
>
>  Your use of Yahoo! Groups is subject to the
><http://docs.yahoo.com/info/terms/>Yahoo! Terms of Service.


--
_______________________________________________________________
Robert P. Futrelle      | Biological Knowledge Laboratory
     Associate Professor  | College of Computer  and Information
                          |     Science MS WVH202
Office: (617)-373-4239  | Northeastern University
Fax:    (617)-373-5121  | 360 Huntington Ave.
futrelle@...    | Boston, MA 02115
           http://www.ccs.neu.edu/home/futrelle
    http://www.bionlp.org   http://www.diagrams.org

            http://biologicalknowledge.com
         mailto:biologicalknowledge@...
_______________________________________________________________

Googling on

automated "music composition" ai

produced about 12,500 page hits.

I think you'll find what you need there,

though I realize there may be someone on this

list who is familiar with the area (I'm not).

("automated" gets more hits than "automatic")

 - Bob Futrelle

Dear everyone:

 

Could anyone please tell me if I would like to do research on Automatic Music Composition Generator, who is the professor in this area and what is the website to focus on?

Thank you for all answers.

 

Best regards,

Jaidee, Mr.

               

---

Yahoo! Music Unlimited - Access over 1 million songs. Try it free.

---

YAHOO! GROUPS LINKS

•  Visit your group "aima-talk" on the web.
   
•  To unsubscribe from this group, send an email to:
   aima-talk-unsubscribe@yahoogroups.com
   
•  Your use of Yahoo! Groups is subject to the Yahoo! Terms of Service.

Hello all,

I met this when I was doing my homework.

Prove that, if the heuristic function h never overestimates by more
than cost c, A* search using h returns a solution whose cost exceeds
that of the optimal solution by no more than c.

This is Question 4.6 from AIMA book. Any hints will be very much
appreciated.

Regards,
odm

hi,
     is it true that CRFs require less training data
than other discriminative models.If yes what is the
reason?

Pranjal Awasthi





__________________________________
Yahoo! Mail - PC Magazine Editors' Choice 2005
http://mail.yahoo.com

Hi everyone,

the book explains pattern databases on page 107 (2nd ed.). It says that it
is currently unknown how how the pattern database approach can be applied
to Rubik's cube. However there is now the work by Korf

@InProceedings{Kor97,
   author = 	 {Richard E. Korf},
   title = 	 {Finding Optimal Solutions to {R}ubik's {C}ube Using
Pattern Databases},
   booktitle = 	 {Proceedings of the 14th National Conference on
Artificial
                   Intelligence and 9th Innovative Applications of
                   Artificial Intelligence Conference},
   pages =  {700-705},
   year =  1997,
   publisher =  {MIT Press}
}

which addresses this question. It would be nice if the authors could
mention this work.

Cheers,
Jan Smaus, Freiburg