【摘要】 本文研究了一种基于遗传算法和BP神经网络的BCI分类器,利用遗传算法和BP神经网络相结合的方式进行网络的学习。把网络的性能作为适应度函数,利用遗传算法快速筛选出合适的神经网络,在利用2003年BCI竞赛的标准数据测试后,可证明该方式可以大大缩短神经网络的训练时间,有效提高神经网络的工作效率,准确率比单纯使用BP神经网络高10%左右。更多还原

【Abstract】 One BCI classifier has been analyzed in this paper, which combines genetic algorithm (GA) with BP neural network to train network’s weights and biases. During genetic evolution, networks’ effectiveness was used as fitness function to filter a much better network. By this means, training time is largely shortened. The final mixed classifier’s recognizing rate can be increased by about 10% than neural-network-alone-based classifier.更多还原