Effort associated with a class of random optimization methods (Q2640454)

When differentiability is not assumed, random methods can adequately be used as a way to optimize a given function. The question that then most naturally may arise is that of the required effort to arrive at an acceptable solution. The author analyses this problem in the case of a given class of random algorithms. Two types of efforts are distinguished: the number of iterations to arrive at a value near enough to the global minimum and the frequency of improvements of the objective function for a given number of iterations. For both these parameters average values are evaluated and proposed by the author. Although well constructed and written, this short article is of somewhat limited interest due to its specialized field, further reduced by the given class of algorithms to which it applies. However, for the reader confronted with this very problem, this paper will certainly provide an inspiring source of information and for this reason can be considered as a valuable contribution.