Behavior of Evolutionary Algorithms in Chaotically Changing Fitness Landscapes

We study an evolutionary algorithm used for optimizing in a chaotically changing dynamic environment. The corresponding chaotic non–stationary fitness landscape can be characterized by quantifiers of the underlying dynamics–generating system. We give experimental results about how these quantifiers, namely the Lyapunov exponents, together with the environmental change period of the landscape influence performance measures of the evolutionary algorithm used for tracking optima.

• Type: Conference
• Year: 2004
• Where: PPSN
• Authors: Hendrik Richter
• Keywords: Distributed And Parallel Computing, Evolutionary Algorithm, Landscape Influence Performance, Non–stationary fitness Landscape, PPSN 2004