论文部分内容阅读
针对最小二乘支持向量机用于在线建模时存在的计算复杂性问题,提出一种动态无偏最小二乘支持向量回归模型.该模型通过改进标准最小二乘支持向量机结构风险的形式消除了偏置项,得到了无偏的最小二乘支持向量机,简化了回归系数的求解.根据模型动态变化过程中核函数矩阵的特点,设计了基于Cholesky分解的在线学习算法.该算法能充分利用历史训练结果,减少计算复杂性.仿真实验表明了所提出模型的有效性.
Aiming at the computational complexity of LS-SVM for on-line modeling, a dynamic unbiased least squares support vector regression model is proposed, which improves the standard least squares support vector machine (SVM) by eliminating the structural risk The biased term is obtained and an unbiased least square support vector machine is obtained, which simplifies the solution of the regression coefficients.According to the characteristics of the kernel function matrix in the dynamic process of the model, an online learning algorithm based on Cholesky decomposition is designed, which can make full use of Historical training results, reducing the computational complexity.The simulation results show the effectiveness of the proposed model.