Graph-based normalization and whitening for non-linear data analysis

In this paper we construct a graph-based normalization algorithm for non-linear data analysis. The principle of this algorithm is to get a spherical average neighborhood with unit radius. First we present a class of global dispersion measures used for “global normalization”; we then adapt these meas...

Celý popis

Uloženo v:
Podrobná bibliografie
Vydáno v:Neural networks Ročník 19; číslo 6; s. 864 - 876
Hlavní autor: Aaron, Catherine
Médium: Journal Article Konferenční příspěvek
Jazyk:angličtina
Vydáno: Oxford Elsevier Ltd 01.07.2006
Elsevier Science
Elsevier
Témata:
Algorithms
Applied sciences
Artificial Intelligence
Cluster Analysis
Computer Graphics
Computer science; control theory; systems
Connectionism. Neural networks
Exact sciences and technology
Geodesic distance
Graph
Humans
Mathematics
Models, Statistical
Nonlinear Dynamics
Normalization
Other Statistics
SOM
Statistics
SOM
Normalization
Graph
Geodesic distance
Data analysis
Procedure call
Multigraph
Geodesic structure
Neural network
Algorithm
Weighted graph
Self-organising feature maps
Geodesic
Analogy
Dispersion measure
normalization
graph
geodesic distance
ISSN:0893-6080, 1879-2782
On-line přístup:Získat plný text
Tagy: Přidat tag
Žádné tagy, Buďte první, kdo vytvoří štítek k tomuto záznamu!
Popis
Shrnutí:In this paper we construct a graph-based normalization algorithm for non-linear data analysis. The principle of this algorithm is to get a spherical average neighborhood with unit radius. First we present a class of global dispersion measures used for “global normalization”; we then adapt these measures using a weighted graph to build a local normalization called “graph-based” normalization. Then we give details of the graph-based normalization algorithm and illustrate some results. In the second part we present a graph-based whitening algorithm built by analogy between the “global” and the “local” problem.
Bibliografie:ObjectType-Article-2
SourceType-Scholarly Journals-1
ObjectType-Feature-1
content type line 23
ObjectType-Article-1
ObjectType-Feature-2
ISSN:0893-6080
1879-2782
DOI:10.1016/j.neunet.2006.05.022