Abstract: Hierarchical discriminant regression (HDR) casts classification problems (class labels as output) and regression problems (numeric values as output) into a unified regression problem. Clustering is performed in both output space and input space at each internal node, termed “doubly clustered” and discriminants in the input space are automatically derived from the clusters in the input space. A hierarchical probability distribution model is applied to the resulting discriminating subspace at each internal node. This realizes a coarse-to-fine approximation of probability distribution of the input samples. It is helpful in high-dimension data retrieval. Kernel method on clustering in input space is used, so the impact of the nonlinear border can be effectively reduced and the results of the retrieval will be more accurate.

Key words: Kernel, Discriminant regression, Hierarchical discriminant regression (HDR)

摘要： 层次判别回归把分类和回归统一成回归问题，在每个内部节点进行输入和输出空间的双聚类，输入空间的分类子空间可以自动从聚类中获得，再用层次概率分布模型计算判别子空间，构成输入空间由粗到精的概率分布，可以准确而且快速地实现高维数据的检索。该文提出了利用核函数在输入空间先对样本进行核聚类，就能够有效降低非线性分类边界的影响，使得检索结果更加准确。

关键词: 核函数, 判别回归, 层次判别回归

CLC Number: 

• TP391.41