A model selection algorithm for the Generative Gaussian Graph

Gaillard Pierre; Michael Jean-Marie Aupetit

A model selection algorithm for the Generative Gaussian Graph

Gaillard Pierre, Michael Jean-Marie Aupetit

Commissariat à l’énergie atomique et aux énergies alternatives

Research output: Contribution to conference › Paper › peer-review

Abstract

One way to learn the topology of principal manifolds known only through a finite set of points is to locate some prototypes representative of the set of points in the ambient space, and to build a graph connecting them. Then the topological characteristics of interest can be extracted from the resulting graph. This paper studies the GenerativeGaussian Graph proposed by Aupetit that allows modelling the connectedness of some principal manifolds. In the practical implementation of this algorithm, two parameters are required to be chosen (the number of prototypes and a threshold to prune the graph). The goal of this pa-per is to develop an automatic method for choosing the values these parameters.

Original language	English
Publication status	Published - 2009
Externally published	Yes

Access to Document

https://scholar.google.com/citations?view_op=view_citation&hl=en&user=UszzuaAAAAAJ&cstart=100&pagesize=100&citation_for_view=UszzuaAAAAAJ:UebtZRa9Y70C

Cite this

@conference{749eb4c3faec4dfdbcf628692c8100f9,

title = "A model selection algorithm for the Generative Gaussian Graph",

abstract = "One way to learn the topology of principal manifolds known only through a finite set of points is to locate some prototypes representative of the set of points in the ambient space, and to build a graph connecting them. Then the topological characteristics of interest can be extracted from the resulting graph. This paper studies the GenerativeGaussian Graph proposed by Aupetit that allows modelling the connectedness of some principal manifolds. In the practical implementation of this algorithm, two parameters are required to be chosen (the number of prototypes and a threshold to prune the graph). The goal of this pa-per is to develop an automatic method for choosing the values these parameters. ",

author = "Gaillard Pierre and Aupetit, {Michael Jean-Marie}",

year = "2009",

language = "English",

}

TY - CONF

T1 - A model selection algorithm for the Generative Gaussian Graph

AU - Pierre, Gaillard

AU - Aupetit, Michael Jean-Marie

PY - 2009

Y1 - 2009

N2 - One way to learn the topology of principal manifolds known only through a finite set of points is to locate some prototypes representative of the set of points in the ambient space, and to build a graph connecting them. Then the topological characteristics of interest can be extracted from the resulting graph. This paper studies the GenerativeGaussian Graph proposed by Aupetit that allows modelling the connectedness of some principal manifolds. In the practical implementation of this algorithm, two parameters are required to be chosen (the number of prototypes and a threshold to prune the graph). The goal of this pa-per is to develop an automatic method for choosing the values these parameters.

AB - One way to learn the topology of principal manifolds known only through a finite set of points is to locate some prototypes representative of the set of points in the ambient space, and to build a graph connecting them. Then the topological characteristics of interest can be extracted from the resulting graph. This paper studies the GenerativeGaussian Graph proposed by Aupetit that allows modelling the connectedness of some principal manifolds. In the practical implementation of this algorithm, two parameters are required to be chosen (the number of prototypes and a threshold to prune the graph). The goal of this pa-per is to develop an automatic method for choosing the values these parameters.

M3 - Paper

ER -

A model selection algorithm for the Generative Gaussian Graph

Abstract

Access to Document

Fingerprint

Cite this