DANCo: An intrinsic dimensionality estimator exploiting angle and norm concentration

作者：

Highlights：

• We propose a family of novel intrinsic dimensionality estimators.

• Our methods reduce the underestimation problem affecting most of the estimators.

• Our intensive experimental tests show the quality of the proposed methods.

摘要

Highlights•We propose a family of novel intrinsic dimensionality estimators.•Our methods reduce the underestimation problem affecting most of the estimators.•Our intensive experimental tests show the quality of the proposed methods.

论文关键词：Intrinsic dimensionality estimation,Manifold learning,Von Mises distribution,Nearest neighbor distance distribution,Kullback–Leibler divergence

论文评审过程：Received 13 July 2013, Revised 24 November 2013, Accepted 19 February 2014, Available online 5 March 2014.

论文官网地址：https://doi.org/10.1016/j.patcog.2014.02.013