As I said, it really doesn't matter. All you need to know is that there is a value being maximized, determined by the behavior of the actions generated by the boolean value returned at a given time step for this 'program' that I am evolving through these conjugated functions.

Quote: Original post by visage
Any thoughts on altering the sub block data to help conversion?

Maybe have 'mutations' only apply algebraically viable replacements. Some examples:
¬¬x -> x
(¬x)∧(¬y) -> ¬(x∨y)
(x ∨ ¬x) -> 1
(x ∧ ¬x) -> 0

Quote: Original post by visage
As I said, it really doesn't matter.

I do respect that you want to keep many of the details a secret, but the fitness function really does matter. It always matters in evolutionary algorithms.

Would you say that your results are very similar to what biologists would call punctuated equalibrium? .. Many generations of little to no improvement and then a single member becomes significantly better, followed by a short period where its genes saturate the population?