Genetic neural network based on adaptive potential well crossover operator and its application in recognition of blue-green algal

In accordance with issues of high randomness, slow convergence speed and prematurity of genetic algorithm, an adaptive potential well crossover operator is proposed. The operator concludes the wave function of particle in square well with infinite depth in quantum mechanics and the concept of quantization of energy. In the algorithm, the operator and the strategy of deterministic crowding mechanism are used in neural network training. It has been demonstrated by simulation results and the pattern recognition experiment on blue-green algal that the approach not only has the properties of high convergence speed and strong searching ability but also has high efficiency in pattern recognition