基于子空间插值的领域适应学习

为了寻找一种更好的目标分类器,提出了基于子空间插值的领域适应学习方法.从源域出发,依次训练中间领域字典,缩小样本分布差,形成一条平滑的虚拟路径,逐步适应目标数据,达到提高识别性能的目的.在CMU-PIE上进行了实验,结果表明此方法较其他领域适应学习方法有更好的分类性能.

有界线性算子空间中最佳逼近的强惟一性

研究了有界线性算子空间中最佳逼近的强惟一性问题 。

Interpolated Tensor Products of Exponential Type Vectors of Unbounded Operators

In this paper we define the tensor products of spaces of exponential type vectors of closed unbounded operators in Banach spaces. Using the real method of interpolation （K-functional） we prove the interpolation theorems that permit to characterize of tensor products of spaces of exponential type vectors, We show an application of abstract results to the theory of regular elliptic operators on bounded domains. For such operators the exponential type vectors are root vectors. Thus we describe the tensor products of root vectors of regular elliptic operators on bounded domains.