LLE(Local linear embedding) is a widely used approach for di mension reduction. The neighborhood selection is an important issue for LLE. In this paper, the ε-distance approach and a slightly modified version of k-nn method are introduced. For different types of datasets, dif ferent approaches are needed in order to enjoy higher chance to obtain better representation. For some datasets w ith complex structure, the proposed ε-distance approach can obtain better representations. Different nei ghborhood selection approaches will be compared by applying them to different kinds of datasets.
Paper
Full text
Neighborhood Function Design for Embedding in Reduced Dimension
Semantic Scholar · Computer Science · 2011
Abstract
LLE(Local linear embedding) is a widely used approach for di mension reduction. The neighborhood selection is an important issue for LLE. In this paper, the ε-distance approach and a slightly modified version of k-nn method are introduced. For different types of datasets, dif ferent approaches are needed in order to enjoy higher chance to obtain better representation. For some datasets w ith complex structure, the proposed ε-distance approach can obtain better representations. Different nei ghborhood selection approaches will be compared by applying them to different kinds of datasets.