基于乘性规则的支持向量域分类器

Support Vector Domain Classifier Based on Multiplicative Updates

该文提出了一种基于支持向量域描述 (SVDD)的学习分类器 .在两类样本分类中 ,该算法在训练时通过对1类样本的描述求取包含 1类样本的球形边界 ,然后通过该边界对两类样本数据进行分类 ,并且在求取边界的优化问题中 ,采用乘性规则来直接求取Lagrange乘子 ,而不是用传统的二次优化方法 .该文所获得的学习算法和支持向量机 (SVM)与序列最小优化 (SMO)算法相比 ,不仅降低了样本的采集代价 ,而且在优化速度上有了很大提高 .通过CBCL人脸库的仿真实验 ,将该算法和SVM、SOM算法的实验结果进行对比 ,说明了该学习算法的有效性 .

This paper proposes a support vector domain classifier based on multiplicative updates. In two-class problem, through description of the training samples from one class, this algorithm obtains the sphere boundary containing these samples, and then uses this boundary to classify the test samples. In addition, instead of the traditional quadratic programming, multiplicative updates is used to solve Lagrange multiplier in optimizing the solution of boundary. Compared to support vector machine (SVM) and sequential minimal optimization (SMO) algorithms, the learning algorithm shown in this paper not only decreases the collecting cost of samples, but also improves greatly the computational speed of optimization. The experiment on CBCL face database illustrates the effectiveness of this algorithm in comparison with SVM and SMO.