[Cs5500] Ahmed was right

[Karl Lieberherr]

Ahmed was right.

A better playground design for MMG is the one he implemented in:
/home/lieber/.www/courses/se-courses/cs5500/f11/software/src/mmg

Define the quality(x,y) = f(x,y) = x*y + (1-x)*(1-y^2).
This will encourage the avatars to find the best y for a given x.

If there is a y so that  x*y + (1-x)*(1-y^2) >= c, it is encouraged
to find such a y.

I see here a form of implication relationship between playgrounds.
We have two playgrounds:

MMG1 = MMG(ForAllExists, claim(c), quality(x,y)=if f(x,y)>=c 1 else 0)
MMG2 = MMG(Survivor, claim(), quality(x,y)=f(x,y))

Avatars that grow up in MMG2 are at least as good as avatars
that grow up in MMG1 when it comes to finding a y maximizing f(x,y).
On the other hand, avatars in MMG1 are more orderly.
If they are good, they converge towards the cmax. IN MMG2
there is no notion of strengthening. All say: "I am the best".

MMG2 is a more general playground. It is not important that
there is a cmax. We are interested in finding a function
g(x) so that f(x,g(x)) is maximized.
x does not have to be a real number but it could be
constraint satisfaction problem.