On the Use of Human-Guided Evolutionary Algorithms for Tackling 2D Packing Problems - Foundations on Natural and Artificial Computation - page 358

Fig. 3. Best solution found in the 29-object instance with the automated EA (left),
and the human-guided EA (right)
the human interaction is focused on augmenting greediness in the initial stages
of the runs, relaxing it in more advances stages of the search. The main goal of
these experiments is determining if human-guidance results in a benefit, rather
than the absolute quality of the results. Fig. 2-4 depict a comparison of the best
solutions found by each of the approaches. As it can be seen, these best solutions
are better in the case of the hgEA. Median values for the aEA are 22, 67 and
68 for each of the three instances, whereas the hgEA provide median values of
20.5, 64 and 64.5 (differences statistically significant at 0.05 level).