Geodesic Learning

Rui Vigelis, Charles Cavalcante

DOI: 10.14209/sbrt.2008.42743

Evento: XXVI Simpósio Brasileiro de Telecomunicações (SBrT2008)

Keywords: Geodesic learning dimensionality reduction data analysis

Abstract

In this paper we develop a new algorithm for estimating the geodesic between points in a point cloud. Geodesics are useful in dimensionality reduction and data analysis. The algorithm exploits data locally, and results in a distance minimizing curve. The convergence was experimentally verified in a toy (swiss roll) and ‘semitoy’ (rendered faces) data set.

Download