[galib] GA program

Peter Jay Salzman p at dirac.org
Mon Nov 19 12:46:42 EST 2007


>> Peter <p at dirac.org> said:     
>>    Hi Somnath,
>>
>>    Please state, clearly and precisely, the problem you want to solve.   For
>>    example:
>>      "I want to find the 1st root of Bessel's function of 2nd order."
>>      "I want to find the 3 cubic roots of unity."
>>      "I want to minimize the gudermanian function."
>>      "I want to find the intersection of tan(x) and tanh(x)."
>>    If you can state your problem very clearly, I will try to help.
>>
>> On Mon 19 Nov 07,  4:40 PM, Somnath Bhattacharjee <sn.bhattacharjee at yahoo.co.in> said:
>    Hi
>    Pete.
>
>    Like your 1st statement, "I want to find 2 roots of
>    the given function (2nd order function)."
>    y=X*X+5*X+6;
>     
>    Regards
>    Somnath


Somnath, you keep saying you want to find the zeros of a function, but
in your code, you blindly call:

   ga.minimize();     

Since your fitness function does nothing but return the value of y(x), a fit
x is one that makes y(x) as small as possible (since you call minimize).  An
unfit x make y(x) large (since you call minimize).

Your first gene's value was the minimum of x^2 + 5x + 6 and I believe your
2nd gene's value was nothing but the starting IC, evolved under mutation and
crossover.

If you want to find roots, you need to write a fitness function that returns
a high fitness when y(x) is close to zero and a low fitness when y(x) is far
from zero.  Then you'd want to maximize your objective.

You *must* read the documentation.  The demo programs are good, but they
won't give you all the info you need.

It sounds like you also need to read a GA tutorial.  I'm not sure you quite
understand the relationship between the objective function and how a GA
arrives at a good candidate solution.

Pete

-- 
GPG Fingerprint: B9F1 6CF3 47C4 7CD8 D33E  70A9 A3B9 1945 67EA 951D
Last night I dreamt of 09-f9-11-02-9d-74-e3-5b-d8-41-56-c5-63-56-88-c0

"A mathematician is a machine for converting coffee    p at dirac.org
 into theorems."     -- Paul Erdös                     http://www.dirac.org



More information about the galib mailing list