We use a population size of P = 10, and the algorithm was allowed to run for a total
of 100 generations. The initial values of the λi variables were chosen from a Gaussian distribution with mean zero and variance one. The initial values of the ηi variables were all set at 3.0; this corresponds to the standard deviation of the Gaussian mutations, so 3.0 means it is possible for a mutation to produce a new λ value up to 3.0 standard deviations away from a parent.