Z. Konfršt (Czech Republic)
Parallel genetic algorithms, Speedup, Super-linearity.
As in many research works of parallel genetic algorithms
(PGAs), claims of a super-linear speedup (super-linearity)
have become so regular that some clariﬁcation is usually
needed. This paper focuses on the estimation of compu
tation characteristics from parallel computing. PGAs are
stochastic based algorithms, so the application rules from
parallel computing is not straightforward. We derive total
(parallel) run times from population sizing, the estimation
of selection intensity and convergence time. The ﬂawless
calculation of total run is essential for obtaining the char
acteristics such as speedup S(.) and others. However, al
though the process of derivation such characteristics is not
simple, it is possible, as it is presented in the paper.