A representation is the format used to encode a candidate solution. Common representations include binary strings, real-valued vectors, permutations, trees, graphs, and neural network parameters.

Fitness evaluation translates a candidate into a numerical score. In simple problems, fitness may equal objective value. In harder problems, it may combine performance, penalties, and constraints.

Selection pressure controls how strongly better solutions are favored. Too little pressure can make progress slow; too much pressure can reduce diversity and cause premature convergence.

For roulette selection, a candidate with fitness $f_i$ may be selected with probability $p_i = \frac{f_i}{\sum_j f_j}$ when all fitness values are nonnegative.