All code is in basic python 3.6 (no NumPy). MaxOnes is a problem in which a population of individuals is evaluated based on their proximity to a target bit string consisting solely of ones. Fitness is defined to be the ratio of correct to incorrect guesses (e.g. target=[1,1], i=[0,1], fitness = 0.5). Mutation, then, corresponds to randomized bit-flipping.

AFPO is a multi-objective evolutionary algorithm which seeks to preserve the innovations of younger cohorts by protecting them from domination by their older counterparts. An individual is only dominated (that is, removed from the population) if another individual has BOTH higher fitness and a lower age. Descendants inherit the age of their parent.

In the event of a tie (same fitness; same age) the individual who was spawned most recently is preserved. The reason for this is to expose a greater number of possible phenotypic pathways to selection (Wagner, 2011).

Parameters like population size and length of genetic strings are modifiable in the "Run" file.

Below is an example fitness plot for AFPO where the length of the target bitstring=500. The population size was set to 100, and the algorithm was terminated after 1000 generations. Normally, we'd expect to see more usurpations by younger cohorts, but given the nature of the problem, innovations are necessarily minor. A single bit-flip won't result in a significant change in fitness. For smaller target strings, there is much greater turnover. However, convergence also happens much faster, so this behavior isn't as apparent as in Cheney et.al 2018.