精确增量式ε型孪生支持向量回归机

Accurate incremental ε-twin support vector regression

中文关键词  机器学习  增量学习  在线学习  孪生支持向量回归机  学习算法  可行性分析  有限收敛性分析

英文关键词  machine learning  incremental learning  online learning  twin support vector regression  learning algorithms  feasibility analysis  finite convergence analysis

中文摘要

      为了解决现有ε型孪生支持向量回归机的训练算法无法高效处理线性回归的增量学习问题, 提出了一种精 确增量式ε型孪生支持向量回归机(AIETSVR). 首先通过计算新增样本的拉格朗日乘子以及调整边界样本的拉格朗 日乘子, 尽可能减少新增样本的二次损失对原有样本的影响, 使得大部分原有样本依然满足Karush–Kuhn–Tucker (KKT)条件, 从而获得一个有效的初始状态; 其次对异常拉格朗日乘子逐步调整至满足KKT条件; 然后从理论上分 析了AIETSVR的可行性和有限收敛性; 最后在基准测试数据集上进行仿真. 结果表明, 与现有的代表性算法相比, AIETSVR能够获得精确解, 在缩短大规模数据集的训练时间上优势显著.

英文摘要

      In ε-twin support vector regression, to solve the problem that the existing algorithms can not efficiently deal with the incremental learning for linear regression, an accurate incremental ε-twin support vector regression (AIETSVR) is proposed. First, by calculating the Lagrangian multiplier of the new sample and adjusting the Lagrangian multipliers of the boundary samples, the influence generated by the quadratic loss of the new sample on the existing samples is minimized. Therefore, most of the existing samples still meet the Karush–Kuhn–Tucker (KKT) conditions, and a valid initial state is obtained. Then, the exceptional Lagrangian multipliers are gradually adjusted to conform to the KKT conditions. Next, the feasibility and finite convergence of AIETSVR are theoretically analyzed. Finally, the simulation is conducted on benchmark datasets. Compared with the existing representative algorithms, the results show that AIETSVR can obtain accurate solutions and has a great advantage in shortening training time for large-scale dataset.