论文部分内容阅读
在K-SVCR算法结构的基础上构造了新的模型.模型的特点是它的一阶最优化条件可以转化为一个线性互补问题,通过Lagrangian隐含数,可以将其进一步转化成一个强凸的无约束优化问题.利用共轭梯度技术对其进行求解,在有限步内得到分类超平面.最后在标准数据集进行了初步试验.试验结果显示了提出的算法在分类的精度和速度上都有明显提高.
A new model is constructed based on the structure of K-SVCR algorithm.The characteristic of the model is that its first-order optimization condition can be transformed into a linear complementary problem, which can be further transformed into a strongly convex by Lagrangian implication The unconstrained optimization problem is solved by the conjugate gradient technique and the classification hyperplane is obtained in a finite step.Finally, a preliminary experiment on a standard dataset is carried out.The experimental results show that the proposed algorithm has both accuracy and speed of classification Significantly improved.