Clusters and factors: neural algorithms for a novel representation of huge and highly multidimensional data sets

Lelu, Alain

doi:10.1007/978-3-642-51175-2_27

Cited by 9 publications

(7 citation statements)

References 5 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Principal-component analysis and correspondence analysis were used to visualize the relationships between the isolates. Classification of the isolates into homogeneous metabolic groups was achieved by use of the KMACL4 software, a new unsupervised classification method for building overlapping clusters of bacteria and determining the main characteristics of these classes (2,27). This software was designed for marrying the benefits coming out of both cluster and factor representations.…”

Section: Methodsmentioning

confidence: 99%

Metabolic and Genotypic Fingerprinting of Fluorescent Pseudomonads Associated with the Douglas Fir-Laccaria bicolor Mycorrhizosphere

Frey¹,

Frey‐Klett²,

Garbaye³

et al. 1997

Appl Environ Microbiol

143

View full text Add to dashboard Cite

A collection of 300 isolates of fluorescent pseudomonads was established from Douglas fir-Laccaria bicolor mycorrhizas and mycorrhizosphere and from adjacent bulk soil. These isolates were first phenotypically characterized with the Biolog method. Taxonomic identification assigned 90% of the isolates to the different biovars of Pseudomonas fluorescens, with inverted frequencies of biovars V and I from the bulk soil to the mycorrhizas, suggesting that the mycorrhizas exert a selective stimulation of the P. fluorescens bv. I and a counterselection of the P. fluorescens bv. V present in the soil. Multivariate analyses of the carbon source utilization data led to the definition of homogenous metabolic groups and to the identification of the most discriminating substrates for each group. The isolates from the mycorrhizosphere and from the mycorrhizas seem to preferentially utilize carbohydrates, in particular trehalose, which is the most abundant carbohydrate accumulated in the mycelium of L. bicolor. The results suggest that L. bicolor exerts a trehalose-mediated selection on the fluorescent pseudomonads present in the vicinity of the mycorrhizas. Isolates of P. fluorescens from the mycorrhizosphere and mycorrhizas were then genotypically characterized by restriction fragment length polymorphism of PCR-amplified 16S rRNA genes and enterobacterial repetitive intergenic consensus-PCR DNA fingerprinting. Both methods revealed a high genetic polymorphism within the population studied, which was well correlated with the phenotypic characterization.

show abstract

Section: Methodsmentioning

confidence: 99%

Metabolic and Genotypic Fingerprinting of Fluorescent Pseudomonads Associated with the Douglas Fir-Laccaria bicolor Mycorrhizosphere

Frey¹,

Frey‐Klett²,

Garbaye³

et al. 1997

Appl Environ Microbiol

143

View full text Add to dashboard Cite

show abstract

“…The problem of optimisation is NP-hard, they can only be made to converge towards a local optimum which depends on their initialisation (for example the initial positions of centres randomly generated or picked out from the data) or on the order of the data. This disqualified them because we set the proviso that the results should be independent from the initial conditions -our method of axial K-means (Lelu, 1994) is part of this family; the local optima for a given number of clusters mostly reveal the same main clusters which are often trivial but can also make the most interesting ones of average or low size appear/disappear/amalgamate or split. Quite a lot of incremental variants of these methods have been proposed (Binztock and Gallinari, 2002;Chen et al, 2003) and a partial review of these can be found in Lin et al (2004).…”

Section: Adapting Methods With Mobile Centres To Incrementalitymentioning

confidence: 99%

“…To measure the links, we chose to use the cosine in the distributional space (Lelu, 2003), a measure linked to the Hellinger distance (Domengès and Volle, 1979) Worthwhile characteristics follow, particularly:…”

Section: Knns Graphmentioning

confidence: 99%

Paving the way for next generation data-stream clustering: towards a unique and statistically valid cluster structure at any time step

Cuxac

Lelu

Cadot

2011

IJDMMM

Self Cite

View full text Add to dashboard Cite

Abstract:In the domain of data-stream clustering, e.g., dynamic text mining as our application domain, our goal is two-fold and a long term one: 1 at each data input, the resulting cluster structure has to be unique, independent of the order the input vectors are presented 2 this structure has to be meaningful for an expert, e.g., not composed of a huge 'catch-all' cluster in a cloud of tiny specific ones, as is often the case with large sparse data tables.The first preliminary condition is satisfied by our Germen density-mode seeking algorithm, but the relevance of the clusters vis-à-vis expert judgment relies on the definition of a data density, relying itself on the type of graph chosen for embedding the similarities between text inputs. Having already demonstrated the dynamic behaviour of Germen algorithm, we focus here on appending a Monte-Carlo method for extracting statistically valid inter-text links, which looks promising applied both to an excerpt of the Pascal bibliographic database, and to the Reuters-RCV1 news test collection. Though not being a central issue here, the time complexity of our algorithms is eventually discussed. 342P. Cuxac et al.

show abstract

“…The method adopted here is an axial k‐means technique (AKM or KMA), developed by one of us and implemented in the software Neuronav (Diatopie). It is a variant of k‐means partitioning methods (Lloyd, 1982, building on an Bells's lab unpublished paper by the same author, 1957; McQueen, 1967), which brings significant improvements (Lelu, 1994, 2008), giving rise to concept‐vectors and document orthogonal and oblique projections on each cluster axod. It is related to the spherical k‐means family (Domengès & Volle, 1979), with a robust theoretical basis.…”

Section: Methodsmentioning

confidence: 99%

Hybrid citation-word representations in science mapping: Portolan charts of research fields?

Zitt¹,

Lelu²,

Bassecoulard³

2010

J. Am. Soc. Inf. Sci.

Self Cite

View full text Add to dashboard Cite

The mapping of scientific fields, based on principles established in the seventies, has recently shown a remarkable development and applications are now booming with progress in computing efficiency. We examine here the convergence of two thematic mapping approaches, citation-based and word-based, which rely on quite different sociological backgrounds. A corpus in the nanoscience field was broken down into research themes, using the same clustering technique on the 2 networks separately. The tool for comparison is the table of intersections of the M clusters (here M = 50) built on either side. A classical visual exploitation of such contingency tables is based on correspondence analysis. We investigate a rearrangement of the intersection table (block modeling), resulting in pseudo-map. The interest of this representation for confronting the two breakdowns is discussed. The amount of convergence found is, in our view, a strong argument in favor of the reliability of bibliometric mapping. However, the outcomes are not convergent at the degree where they can be substituted for each other. Differences highlight the complementarity between approaches based on different networks. In contrast with the strong informetric posture found in recent literature, where lexical and citation markers are considered as miscible tokens, the framework proposed here does not mix the two elements at an early stage, in compliance with their contrasted logic.

show abstract

Clusters and factors: neural algorithms for a novel representation of huge and highly multidimensional data sets

Cited by 9 publications

References 5 publications

Metabolic and Genotypic Fingerprinting of Fluorescent Pseudomonads Associated with the Douglas Fir-Laccaria bicolor Mycorrhizosphere

Metabolic and Genotypic Fingerprinting of Fluorescent Pseudomonads Associated with the Douglas Fir-Laccaria bicolor Mycorrhizosphere

Paving the way for next generation data-stream clustering: towards a unique and statistically valid cluster structure at any time step

Hybrid citation-word representations in science mapping: Portolan charts of research fields?

Contact Info

Product

Resources

About