To compare the algorithms we use pair-wise Student T tests with
95% confidence interval for the means of evaluations to find the
optimum. Due the large number of simulations we assume the
normality of the variables. In the comparisons we use the symbol
‘+’ if the algorithm is better, the symbol ‘-‘ if is worse and ‘=’ if
the algorithms are not statistically different.
For each problem we also compare results with other previously
referred algorithms. However results published for these problems
are not always available with precision (in some cases only
logarithmic graphs are printed).