摘要
利用支持向量机中的拉格朗日系数,从系数所表达的意义出发,根据Kolmogorov算法随机性理论,定义了可信度和可靠性,提出了一种小样本集可信向量机,该向量机在预测对象类别的同时,给出本次预测的可信程度和可靠性,丰富了支持向量机的输出信息。同时通过在预测的过程中有限增加训练集中的有用特征信息,提高了预测的准确率,在只有少量训练样本的情况下具有较好的性能。
A small sample set confidence SVM is put forward based on Kolmogorov＇s algorithmic theory of randomness using the Lagrangian coefficients of support vector machine according to their physical meanings. The confidence and the credibility of prediction are subsequently defined. When predicting,this machine can output the confidence and the credibility as well as the label of testing examples. And correction portion is improved by adding some testing sample whose characters are helpful for machine into the training sample set while prediction.
出处
《计算机与网络》
2009年第13期36-38,43,共4页
Computer & Network
关键词
小样本集可信向量机
可信度
可靠性
随机性理论
small sample set confidence support vector machine
confidence
credibility
algorithmic theory of randomness
