Multiobjective Memetic Estimation of Distribution Algorithm Based on an Incremental Tournament Local Searcher

Algorithm 1

The main loop of RM-MEDA.

Step  1. Initialization: Set 𝑡=0, generate initial population P(0) and evaluate them.

Step  2. Termination: If 𝑡>𝐺_max is satisfied, export P(t) as the output of the algorithm, and stop, else go to step 3.

Step  3. Modeling: Perform local PCA to partition P(t) into 𝐾 disjoint clusters 𝑆¹,𝑆²,…,𝑆^𝐾. For each cluster 𝑆^𝑗,

build model (2) by (4) and (5).

Step  4. Sampling: Sample new population O(t) from model (2) and evaluate O(t).

Step  5. Non-dominated Sorting and Crowding Distance Assignment: Use the famous fast non-dominated sorting procedure

and select solutions at first several low ranks from P(t) and O(t). At the critical rank, the crowding distance computation

is employed to select individuals with higher values of crowding distance. Then, P(t + 1) is created.

Step  6. set 𝑡∶=𝑡+1 and go to Step 2.