摘要
传统机器学习均假定测试域和训练域处于同一概率分布,但现实中往往因各种原因引起所采集到的样本数据可能存在扰动或噪音情况,导致概率密度估计不一定准确。为有效解决这一问题,提出一种新的领域自适应数据集概率密度估计(A-RSDE)算法。该算法可充分学习源域(训练域)概率密度分布知识,使目标域(测试域)概率密度估计更接近真实概率密度分布。实验证明,该算法具有有效性,且实现了数据浓缩的目的。
The traditional machine learning assumes that the testing dataset has the same probability distribu- tion as the training one. However, in reality, the disturbance or noise which may exist in the collected datasets because of various reasons will cause the incorrect probability density estimation. In order to solve the problem, a new probability density estimation of domain adaptation （A-RSDE） is proposed. It gains the source domain＇s （training dataset） probability density knowledge, making the target domain＇s （testing dataset） probability density estimation closer to the true probability density distribution. Tests show that it is effective and achieve the objective of data concentration.
出处
《温州职业技术学院学报》
2014年第4期38-42,59,共6页
Journal of Wenzhou Polytechnic
基金
国家自然科学基金(60903100)
江苏省教育厅高校哲学社会科学研究基金(2012SJB880077)
