摘要
为了解决工程中数据样本较少情况下的准确建模问题,提出了一种集成先验知识的多核线性规划支持向量回归算法.该算法首先通过修改优化目标和不等式约束条件,把来自仿真模型具有偏差的先验知识数据集成到现有的线性规划支持向量回归的学习框架中.然后,引入多核到集成先验知识的线性规划支持向量回归中以实现复杂规律的准确建模.最后,将算法推广到多输入多输出的数据建模中.仿真案例以及在天线和滤波器的实际应用表明:该算法求解简单,具有较好的模型稀疏和准确性.
In order to obtain an accurate model from a small data set,a novel multiple kernel linear programming support vector regression with prior knowledge is presented in this paper.The algorithm firstly incorporates the data that is possibly biased from the prior simulator into the existing linear programming support vector regression by modifying optimization objectives and inequality constraints.Then,multiple kernels are introduced to integrate the linear programming support vector regression with prior knowledge,in order to achieve an accurate modeling for complex and changeful problems.Finally,the algorithm has also been generalized to model the multi-input multi-output data.The synthetic example and practical applications of an antenna and a cavity filter show that the proposed algorithm is simple,and that the obtained model is sparse and accurate.
出处
《自动化学报》
EI
CSCD
北大核心
2011年第3期360-370,共11页
Acta Automatica Sinica
基金
国家重点基础研究发展计划(973计划)(613580205)
国家自然科学基金(51035006
50475171)
中央高校基本科研业务费(JY10000904019)资助~~
关键词
线性规划支持向量回归
先验知识
多核
小样本
天线
滤波器
Linear programming support vector regression (LPSVR)
prior knowledge
multiple kernels
small data set
antenna
cavity filter