CDA: A Novel Clustering Delegate Algorithm Based on Minority Protection

In traditional Gene Expression Programming (GEP), individuals´ survival too much depends on fitness while their relationships are ignored. Borrowing the idea from the minority protection in real life, this study introduces a novel Cluster Delegate algorithm (CDA) and makes the following contributions: (1) propose several new concepts including individual similarity, ß- cluster, and the farthest neighborhood clustering, (2) implement CDA algorithm which clustering population by fitness and selects delegate from each cluster, (3) conduct extensive experiments to show that newly proposed method can accurately discover functions in complex problems.