CAJ | 학술논문

支持向量机对数据的学习往往因为规模过大造成学习困难,增量学习通过把数据集分割成历史样本集和新增样本集,利用历史样本集的几何分布信息,通过定义样本的遗忘因子,提取历史样本集中的那些可能成为支持向量的边界向量进行初始训练.在增量学习过程中对学习样本的知识进行积累,有选择地淘汰学习样本.实验结果表明,该算法在保证学习的精度和推广能力的同时,提高了训练速度,适合于大规模分类和在线学习问题.
지지향량궤대수거적학습왕왕인위규모과대조성학습곤난,증량학습통과파수거집분할성역사양본집화신증양본집,이용역사양본집적궤하분포신식,통과정의양본적유망인자,제취역사양본집중적나사가능성위지지향량적변계향량진행초시훈련.재증량학습과정중대학습양본적지식진행적루,유선택지도태학습양본.실험결과표명,해산법재보증학습적정도화추엄능력적동시,제고료훈련속도,괄합우대규모분류화재선학습문제.
To learn for large scale datasets is difficult using support vector machine.Datasets are divided into histor ydataset and incremental datasets.a new algorithm based on the geometrical knowledge of history samples is presented.Firstly,the border vectors of history samples are selected by redefining forgetting factor of sample,and then SVM is fast trained by these border vectors.Secondly,all samples'knowledge is accumulated and some samples is discarded effectively in the incremental learning process.The numerical experiments on benchmark datasets show that the proposed algorithm is considerably faster than the standard SVM and the classical incremental algorithm.