Manual (40 points)
------

* Check that Population.nextGeneration() sorts the new generation of
 individuals. - 5 points

* Check that Evolution.getBest() delegates to Population.getBest() (as
 opposed to using population's fields directly). - 5 points

* Check that Evolution.evolve() is simple: stops if getBest().isIdeal(), or
 if maximum number of generations reached, otherwise calls population's
 nextGeneration() method. - 15 points.

* Check that SubsetSumIndividual.mutate() and
 SubsetSumIndividual.crossover() look like being implemented correctly - 15 points.


Automatic (60 points)
---------

* Check that KnapsackIndividual does not override clone(), and overrides
 evaluate() - 10 points

* Subclass Individual, and check that evaluate() is called exactly once
 per multiple calls to getFitness() - 5 points.

* Using same subclass, check that isIdeal() returns true iff evaluate
 returned 0 - 5 points

* Subclass SubsetSumIndividual, and create two Individuals that return
 different values in evaluate(). Check that compareTo() is correct on
 freshly created individuals i1 and i2 both for i1.compareTo(i2) and for
 i2.compareTo(i1). Individuals should be freshly created for both
 compareTo()'s! This checks that compareTo() uses getFitness(), and does
 not access the fitness field directly (which may be uninitialized). - 10 points

* Check tournament selection, as follows:

 + subclass SubsetSumIndividual, to gain access to protected evaluate()

 + create 10 SubsetSumIndividual2's (in an array)

 + mutate each 1000 times (ind = ind.mutate()) - this assures that their
   bits are random

 + create TournamentSelection with some probabilities, and call reproduce
   on this "population" 1000 times.

 + each time, check that all individuals in population still return same
   values in evaluate() (checks that clone() deep-copies the booleans
   array), and also check that the returned individuals are instanceof
   SubsetSumIndividual2 (checks that clone() is implemented correctly).

 + each time, also count how many individuals are == to some individual in
   the original population, and check that the final count is approximately
     (1 - p_cross) * (1 - p_mut) - 15 points (3 tests)

* Run the program on a medium size test (20 times) and check that:
+ the program does not throw any exception
+ the output is different on some of the Runs.
15 points