基于修剪技术分级学习的动态模糊神经网络算法研究

摘要

在D-FNN中采用了修剪技术,可以检测到不活跃的模糊规则并加以剔除,从而获得更为紧凑的结构.在D-FNN中,前提参数是在学习过程中自适应地进行调整.由于分级学习策略的应用,大大提高了学习的有效性,加之参数调整只限于线性参数,没有迭代学习,因而学习速度很快,这使得本算法应用于实时学习和控制成为可能.最后针对实际案例进行了仿真分析,验证了该算法的有效性和高效性.%Using a pruning technique can detect non-active fuzzy rules and to be eliminated and the availability of more compact structure in the D-FNN ,in the premise parameters are in the process of adaptive learning to adjust.The way the application of learning strategies, greatly improved the effectiveness of learning.Combined with linear parameter,limited parameter adjustment,there was no iterative learning, and thus learning very fast,which made the algorithm used in real-time learning and control with the possible.Finally, actual case, the simulation analysis showes the algorithm's effectiveness and efficiency.