Abstract: A novel genetic algorithm with dynamic regional multi-species is proposed in this paper. Each of those genetic species occupies a dynamic region that is determined by the weight vector of a fuzzy adaptive Hamming neural network. Through learning and classification of genetic individuals in the evolutionary procedure, the neural network distributes multi-species into different regions of the search space. Furthermore, the neural network dynamically expands each search region or establishes new region for good offspring individuals to continuously keep the diversification of genetic population. As a result, the premature problem inherent in genetic algorithms is alleviated and better tradeoff between the ability of exploration and exploitation can be obtained. The experimental results on the optimization computation of typical multi-modal functions also have shown good performance of the proposed genetic algorithm.