摘要
最大均值差异忽略了单个样本对全局度量贡献的差异性.为此,提出一种最大分布加权均值差异度量方法,采用白化余弦相似性度量为源域和目标域的所有样本设计相应的分布权重,使得每个样本的分布差异信息在全局度量中均得以体现.进一步,结合联合分布调整思想,提出一种基于最大分布加权均值嵌入的领域适应学习算法.实验结果表明,与典型的迁移学习和无迁移学习算法相比,所提出算法在不同类型跨领域图片数据集上均具有较高的分类精度.
Maximum mean discrepancy neglects the difference of contribution of each sample to the global measure.Therefore, a kind of maximum distribution weighted mean discrepancy（MDWMD） methed is proposed, where the whitened cosine similarity is used to design distribution weights for all samples from the source and the target domains. Further,based on the idea of joint distribution adaptation, a domain adaptation learning algorighm based on MDWMD is proposed.Experimental results show that, compared with the typical transfer learning and non-transfer learning algorithms, the proposed algorithm has higher classification accuracy on different types of cross-domain image datasets.
出处
《控制与决策》
EI
CSCD
北大核心
2016年第11期2083-2089,共7页
Control and Decision
基金
国家自然科学基金项目(61273143
61472424)
中央高校基本科研业务费专项资金项目(2013RC10
2013RC12
2014YC07)
关键词
领域适应学习
最大均值差异
分布权重
白化余弦相似性
联合分布调整
domain adaptation learning
maximum mean discrepancy
distribution weight
whitened cosine similarity
joint probaliality adaptation
