The $(1+(\lambda,\lambda))$ Global SEMO Algorithm

The theory of evolutionary algorithms (EAs) for a long time has accompanied our attempts to understand the working principles of evolutionary computation [NW10, AD11, Jan13, DN20]. In the recent years, this field has not only explained existing approaches, but also proposed new operators and algorithms. The theory of multi-objective EAs, due to the higher complexity of these algorithms, is still lagging behind its single-objective counterpart. There are several runtime analyses for various multi-objective EAs which explain their working principles. Also, some new ideas specific to multi-objective evolutionary algorithms (MOEAs) have been developed recently. However, many recent developments in single-objective EA theory have not been exploited in multi-objective evolutionary computation (see Section 2 for more details). In this work, we try to profit in multi-objective evolutionary computation from the ideas underlying the (1 + (λ, λ)) GA. The (1 + (λ, λ)) GA, proposed first in [DDE15], tries to combine a larger radius of exploration with traditional greedy-style exploitation of already detected profitable solutions.