论文部分内容阅读
核函数的参数严惩影响RVM的综合性能。为求得稀疏解、避免过拟合,提出使用遗传算法针对问题背景自动优化核函数的参数。在适应度函数评判下,种群经过选择、交叉和变异迭代进化,高效率地得到最优解,在定义RVM回归性能综合评判批准Fitness作为适应度函数的基础上,使用Matlab遗传算法工具箱和改进的Tipping程序获取sinc数据最优核函数参数,实验证明遗传算法可以高效准备地优化RVM核参数,特别对于具有较多参数的核函数更具实用性。
Kernel parameters severely punished the overall performance of the RVM. In order to obtain the sparse solution and avoid over fitting, a genetic algorithm is proposed to automatically optimize the parameters of the kernel function against the background of the problem. Under the judgment of fitness function, the population evolves iteratively through selection, crossover and mutation to obtain the optimal solution with high efficiency. On the basis of defining fitness as a fitness function by comprehensive evaluation of RVM regression performance, The Tipping program obtains the optimal kernel function parameters of sinc data. Experiments show that the genetic algorithm can optimize the RVM kernel parameters efficiently and effectively, especially for kernel functions with more parameters.