Optimal manifold neighborhood and kernel width for robust non-linear dimensionality reduction

作者：

Highlights：

• Tap the potential of aligned tangent space to find optimal neighborhood.

• Define kernel bandwidth as a function of neighborhood density and linear region.

• Minimize bias and variance error in graph Laplacian.

摘要

•Tap the potential of aligned tangent space to find optimal neighborhood.•Define kernel bandwidth as a function of neighborhood density and linear region.•Minimize bias and variance error in graph Laplacian.

论文关键词：Semi-supervised,Manifold learning,Tangent alignment,Parzen window,Non-linear,Dimensionality reduction

论文评审过程：Received 15 March 2019, Revised 9 July 2019, Accepted 13 August 2019, Available online 19 August 2019, Version of Record 25 October 2019.

论文官网地址：https://doi.org/10.1016/j.knosys.2019.104953