One armed bandit
Two armed bandit problem

Simple local search + a population

init P = {X_1, X_2, X_3, ...X_n}   // population

compute fitness of P
while (best(P) is not goodenough) {
    parent = select(P)             // selection (k armed bandit)
    child = mutate(parent)         // reproduction and variation
    if (fitness(child) > worst(P)) worst(P) <- child   // selection (compete)
}

new innovation --> sweep the population
enemy: population converges!  friend: diversity


STEADY STATE GENETIC ALGORITHM

init P = {X_1, X_2, X_3, ...X_n}   // population
compute fitness of P

while (best(P) is not goodenough) {
    with (xoverprob) {
        (parent1, parent2) = select(P)      // selection (k armed bandit)
        child = crossover(parent1, parent2)   // needs to be smart
        child = mutate(child)       
    }
    else {
         parent = select(P)             // selection (k armed bandit)
         child = mutate(parent)       
    }
    insert(P, child)
}