适用于智能传感器系统的SVM集成研究  

Research on SVM integration for intelligent sensor system

作　　者：卞桂龙[1] 丁毅 沈海斌[1] 

机构地区：[1]浙江大学超大规模集成电路设计研究所,浙江杭州310027 [2]西湖电子集团有限公司,浙江杭州310012

出　　处：《传感器与微系统》2014年第8期44-47,51,共5页Transducer and Microsystem Technologies

摘　　要：以支持向量机(SVM)为代表的人工智能技术在智能传感器系统中得到了广泛的应用,但传统的SVM有"灾难性遗忘"现象,即会遗忘以前学过的知识,并且不能增量学习新的数据,这已无法满足智能传感器系统实时性的要求。而Learn++算法能够增量地学习新来的数据,即使新来数据属于新的类,也不会遗忘已经学习到的旧知识。为了解决上述问题,提出了一种基于壳向量算法的Learn++集成方法。实验结果表明:该算法不但具有增量学习的能力,而且在保证分类精度的同时,提高了训练速度,减小了存储规模,可以满足当下智能传感器系统在线学习的需求。Support vector machine （SVM） as the representative of artificial intelligent techniques has been widely used in the intelligent sensor system, however, traditional SVM suffers from the catastrophic forgetting phenomenon, which results in loss of previously learned information, so it is unable to meet the requirements of real-time intelligent sensor system. The strength of Learn＋＋ lies in its ability to learn new data without forgetting previously acquired knowledge, even when the new data introduce new classes. In order to solve the above problem,a Learn ＋＋ integration method based on hull vectors is proposed. Experimental results show that the algorithm not only has the ability of incremental learning, improve training speed and reduce the storage size, but also can ensure the classification precision ,which meets the current demand of intelligent sensor systems for online learning.

关 键 词：传感器 支持向量机 壳向量 Learn++算法 增量学习 

分 类 号：TP212[自动化与计算机技术—检测技术与自动化装置] TP18[自动化与计算机技术—控制科学与工程]