Grefenstette  went one stage further and used a GA to optimize these parameters for a test bed of problems. He concluded that
resulted in the best performance when the average fitness of each generation was used as the indicator, while
gave rise to the best performance when the fitness of the best individual member in each generation was monitored. The latter is, of course, the more usual performance measure for optimization routines.
In general, the population size should be no smaller than 25 or 30 whatever the problem being tackled, and for problems of high dimensionality larger populations (of the order of hundreds) are appropriate.