The Explo2 Algorithm
Maurice Clerc · 2019
This chapter presents eight profiles and shows how Explo2 tries to “follow” them during the search process. It should be noted that all these profiles begin with N0 null values, since it is recognized that the first random samplings are of exploration, although this is not always correct, because it is indeed possible that a point randomly drawn be inside the exploitation domain of another, even if the probability of this example is generally negligible. Hence, some profiles present a discontinuity at the beginning that Explo2 cannot perfectly negotiate. Comparisons of cumulative distribution functions (CDFs) are objective criteria, but as seen, they can lead to undecidable situations or, at least, be difficult to interpret when curves intersect. On the contrary, if introducing subjectivity in the form of a curve of user-defined quality for the different results obtained, one can then represent each CDF using a single numeric value, which allows a total ordering relation.