How to find the best parameters for a genetic algorithm?

Question

Some structures of the genetic algorithm, such as http://www.aforgenet.com/ , require many parameters, such as mutation speed, population size, etc.

Are there universal best numbers for such parameters? I believe that this depends on the problem (delay in the health function, delay in mutation, delay in recombination, rate of evolution, etc.). My first thought was to use GA to tune another GA.

Any better ideas?

Answer 1

, , , , , , GA . , , , .

Answer 2

, , .

, , , , , "" , , , .

- ( ):
_{(: Scientific-computing.com)}
:

:

:

• : , , , .
• : .

:

• : , "" , .
• : .

. . , GA .

Answer 3

, , , . , , , .

, No Free Lunch Theor , , :

:

• ( - GA)
• ( - - GA)
  • Self Adaptive

, , , , . , , .

, "" :

:

• sweep ( )
• Meta-GA ontop GA
• , .
• Meta-GA Racing

:

, , , . !

Answer 4

.

? - No Free Lunch. , , .

, , . . .

GA GA . GAGA? ...?

Answer 5

. , . GA GA - GA ? , .

, , , .. , , , .

, - , .

Answer 6

, . 0,7-0,9, - 0,1-0,3, . , . GA, .

GA GA, , , , GA? , , , , , - . . , GA , . GA. .

, .
, .