基于KKT条件与壳向量的增量学习算法研究

摘要针对经典支持向量机难以快速有效地进行增量学习的缺点，提出了基于KKT条件与壳向量的增量学习算法，该算法首先选择包含所有支持向量的壳向量，利用KKT条件淘汰新增样本中无用样本，减小参与训练的样本数目，然后在新的训练集中快速训练支持向量机进行增量学习。将该算法应用于UCI数据集和电路板故障分类识别，实验结果表明，该算法不仅能保证学习机器的精度和良好的推广能力，而且其学习速度比经典的SMO算法快，可以进行增量学习。

Research of Incremental Learning Algorithm Based on KKT Conditions and Hull Vectors

Because the classical support vector machine is difficult to realize incremental learning flectly and rapidly when the number of training samples gets larger, this thesis proposed an incremental learning algorithm based on KKT conditions and hull vectors. This algorithm first selects the hull vectors which contain all support vectors. Next, it eliminates the useless samples among newly-added ones by using KKh conditions in order to reduce the number of training samples, then starts increment learning. The experimental results show that this algorithm not only guarantees the precision and good generalization ability of the learning machine, but also faster than the classical SVM algorithm.Therefore, it can be used in incremental learning.