An Indicator and Decomposition Based Steady-State Evolutionary Algorithm for Many-Objective Optimization

Figure 2

Illustration of the objective space partition strategy. It contains two situations. (a) a new individual is generated without updating the ideal point. The distribution of in the corresponding subspaces remains unchanged. We only locate the subspace where the new individual belongs to. Clustering the combined population is unnecessary. (b) A new individual changes the position of the ideal point. The distribution of the combined population must adopt the objective space partition strategy to repartition the objective space into subspaces.

(a) Cluster an offspring

(b) Cluster the combined candidate solutions