合肥工业大学学报（自然科学版）
閤肥工業大學學報（自然科學版）
합비공업대학학보（자연과학판）
JOURNAL OF HEFEI UNIVERSITY OF TECHNOLOGY(NATURAL SCIENCE)
2014年
7期
794-797,844
,共5页
领域适应%迁移学习%实例重构%多分类器集成
領域適應%遷移學習%實例重構%多分類器集成
영역괄응%천이학습%실례중구%다분류기집성
domain adaption%transfer learning%instance reconfiguration%multiple classifiers ensemble
已有的领域适应方法可分为基于实例和基于特征2类，文章在多领域集成框架下提出M A IR方法，以共现特征为桥梁，从原始领域中选取部分实例对目标领域实例进行重构和预标记，并将预标记实例加入到原始领域进行迭代训练。实验结果表明，M AIR算法在分类性能和时间性能上具有较大优势。
已有的領域適應方法可分為基于實例和基于特徵2類，文章在多領域集成框架下提齣M A IR方法，以共現特徵為橋樑，從原始領域中選取部分實例對目標領域實例進行重構和預標記，併將預標記實例加入到原始領域進行迭代訓練。實驗結果錶明，M AIR算法在分類性能和時間性能上具有較大優勢。
이유적영역괄응방법가분위기우실례화기우특정2류，문장재다영역집성광가하제출M A IR방법，이공현특정위교량，종원시영역중선취부분실례대목표영역실례진행중구화예표기，병장예표기실례가입도원시영역진행질대훈련。실험결과표명，M AIR산법재분류성능화시간성능상구유교대우세。
The existing domain adaption approaches can be divided into two types ,i .e .the ones based on the instances and the ones based on the features .Based on the multi-domain ensemble learning ,the multi-domain adaption based on instance reconfiguration (MAIR) is proposed ,in which the common feature is regarded as the bridge ,and parts of instances are selected instead of all instances from source domain to reconfigure and pre-label the instances in target domain .And those pre-labeled in-stances are added into source domain to conduct iterative training .The experimental results show that the proposed method outperforms alternative methods on the task of multi-domain adaption in the ac-curacy of classification and the cost of time .
