Traditional Gene Expression Programming (GEP) can not discover all optimum solutions for specific multimodal function. To solve this problem, the niche technology was applied to GEP. The limitation of the existing optimization for multimodal function based on GEP was analyzed. A modification strategy with niche radius named AMNR (Adaptive Modification with Niche Radius) and an algorithm named NGEP-MFO (Multimodal Function Optimization based on Niche GEP) were proposed. The application domain of traditional GEP was extended. Experiments showed that compared with traditional GEP, NGEP-MFO can improve the successful ration and accuracy greatly when identifying all optimum solutions.
