Re: Probability and Fuzzy Logic

new topic     » topic index » view thread      » older message » newer message

>
>>Michael Sabal wrote:
>>>
>>
>><snip>
>>
>>>
>>> [Fuzzy logic]
>>>
>
>>>Mathematically, if you flip
>>>a coin 10 times and they are all heads, the probability of the next
>>>flip being heads is 50%.  However, we know from experience that the
>>>odds of heads coming up are closer to 25%, getting lower with each
>>>subsequent head.
>>
>><more snip>
>
>
>>Your odd of coin comming up head or
>>tails has NOTHING to do with prior results. You sound like your going
>>with what intuition seems to say in an area where intuition is often
>>wrong. this is probably coloring your memory.
>
This last is correct.  The flipping of a coin is independent of previous
results.  However, the likelihood of having flipped that many heads AND
flipping another does go down.  The formula for that is:

p=((n!*(n-x)!/x!)(.5)^n

Where n=number of flips, and x=number of heads (or tails).


As for fuzzy logic.  It's designed to show POSSIBILITY, not probability.
Kind of a backwards way of looking at it.  There are some very
interesting applications in industry.

The most common example used is simulating balancing a stick stright up.
You can imagine doing this yourself, moving your hand to keep it under
the stick, and keeping it from falling.  You have to make many
decisions, as to how much acceleration with which to move your hand.

It's possible write a program to do all the math, which involves a lot
of trig and differential equations.  Very messy.  Much easier to set up
a table which takes the angle of the stick and comes up with an answer
as to how fast to move the hand.  It's a little more complicated than
this, but still much easier than through brute force.

The beauty, is that you can take qualitative statements from an
expericenced human operator, and use them to guide a computer.  coupled
with the new work on genetic algorithms, this can be extrememly
powerful.

Although, I don't think that it could be used quite as easily for photo
identification.  It would seem like you'd need to be able to recognize
patterns, rather that individual pixels.  Although, I was just reading
an article in BuisinessWeek about new chips designed for visual
applications, so we may not be that far off any more...




______________________________________________________
Get Your Private, Free Email at http://www.hotmail.com

new topic     » topic index » view thread      » older message » newer message

Search



Quick Links

User menu

Not signed in.

Misc Menu