论文部分内容阅读
CMAC 算法中,研究泛化性能是其中一项主要内容.泛化性能好,则网络的学习精度高.本文阐述网络的原理、结构、学习算法,对影响泛化性能的量化精度、采样精度及其之间的关系进行理论分析.并通过计算机仿真验证了当量化精度等于采样精度、量化精度大于采样精度时对网络精度的影响,得出量化精度应该大于采样精度的结论.提出一种利用基于多目标的遗传算法来确定泛化常数和量化精度的方法,并通过实例验证方法的正确性.
CMAC algorithm, the study of generalization performance is one of the main content of the generalized performance, the network learning accuracy.This paper describes the network principles, structure, learning algorithms, the impact of generalized performance of the quantization accuracy, sampling accuracy and And the relationship between them is analyzed theoretically.The simulation results show that when the quantization accuracy is equal to the sampling precision and the quantization accuracy is greater than the sampling accuracy, the conclusion is drawn that the quantization precision should be greater than the sampling precision.An algorithm based on Multi-objective genetic algorithm to determine the generalization and quantization accuracy of the method, and through examples to verify the correctness of the method.