【摘要】 介绍一种稀疏的贝叶斯学习算法———关联向量机(RVM),它在再生核希尔伯特空间中学习,利用贝叶斯方法推理,推广能力好,与支持向量机相比不仅解更为稀疏而且不需要调整超参数。应用RVM的对小样本的良好分类能力,提出一种基于RVM的入侵检测原型系统。更多还原

【Abstract】 A general Bayesian framework for obtaining sparse solutions to the regression and classification tasks utilizing models linear in the parameters is presented. By using a probabilistic Bayesian learning framework an accurate RVM (relevance vector machine) prediction models using fewer basis functions than a comparable SVM (support vector machine) and offering a number of additional advantages is obtained. The generalization ability of the current ID (intrusion detection system) is poor with less priori knowledge. Using RVM in the intrusion detection the generalization ability of the IDS is good when the sample size is small.更多还原