Guys,
This is very confusing.
All the texts on genetic algorithms start with something like:
Lets use a GA to find a minimum or maximum of f(x) in integers.
Then they encode the x (the problem domain) into a binary chromosome.
For example, a 3 bit chromosome would look like 111 (domain value 7).
Now they say, "apply mutation". Ok, suppose the first bit gets
mutated, and chromosome turns to be 011 (value 3).
What has happened that small change in representation has resulted in
a huge change in actual value.
Aren't all the texts simply wrong? Don't we have to think of a
representation where small changes in repr, make small changes in
actual value?
If the mutation makes huge changes in the actual value, we get some
kind of a random search. And where does the "evolution" come in play?
Sincerely,
Stan
|
