Data-driven clustering : new methods and applications

Abstract : This work is concerned with the development and application of novel unsupervised learning methods, having in mind two target applications: the analysis of forensic case data and the classification of remote sensing images. First, a method based on a symbolic optimization of the inter-sample distance measure is proposed to improve the flexibility of spectral clustering algorithms, and applied to the problem of forensic case data. This distance is optimized using a loss function related to the preservation of neighborhood structure between the input space and the space of principal components, and solutions are found using genetic programming. Results are compared to a variety of state-of--the-art clustering algorithms. Subsequently, a new large-scale clustering method based on a joint optimization of feature extraction and classification is proposed and applied to various databases, including two hyperspectral remote sensing images. The algorithm makes uses of a functional model (e.g., a neural network) for clustering which is trained by stochastic gradient descent. Results indicate that such a technique can easily scale to huge databases, can avoid the so-called out-of-sample problem, and can compete with or even outperform existing clustering algorithms on both artificial data and real remote sensing images. This is verified on small databases as well as very large problems. Résumé : Ce travail de recherche porte sur le développement et l'application de méthodes d'apprentissage dites non supervisées. Les applications visées par ces méthodes sont l'analyse de données forensiques et la classification d'images hyperspectrales en télédétection. Dans un premier temps, une méthodologie de classification non supervisée fondée sur l'optimisation symbolique d'une mesure de distance inter-échantillons est proposée. Cette mesure est obtenue en optimisant une fonction de coût reliée à la préservation de la structure de voisinage d'un point entre l'espace des variables initiales et l'espace des composantes principales. Cette méthode est appliquée à l'analyse de données forensiques et comparée à un éventail de méthodes déjà existantes. En second lieu, une méthode fondée sur une optimisation conjointe des tâches de sélection de variables et de classification est implémentée dans un réseau de neurones et appliquée à diverses bases de données, dont deux images hyperspectrales. Le réseau de neurones est entraîné à l'aide d'un algorithme de gradient stochastique, ce qui rend cette technique applicable à des images de très haute résolution. Les résultats de l'application de cette dernière montrent que l'utilisation d'une telle technique permet de classifier de très grandes bases de données sans difficulté et donne des résultats avantageusement comparables aux méthodes existantes.

Machine learning analysis and modeling of interest rate curves

The present research deals with the review of the analysis and modeling of Swiss franc interest rate curves (IRC) by using unsupervised (SOM, Gaussian Mixtures) and supervised machine (MLP) learning algorithms. IRC are considered as objects embedded into different feature spaces: maturities; maturity-date, parameters of Nelson-Siegel model (NSM). Analysis of NSM parameters and their temporal and clustering structures helps to understand the relevance of model and its potential use for the forecasting. Mapping of IRC in a maturity-date feature space is presented and analyzed for the visualization and forecasting purposes.

Using spatially distributed parameters and multi-response objective functions to solve parameterization of complex applications of semi-distributed hydrological models

Application of semi-distributed hydrological models to large, heterogeneous watersheds deals with several problems. On one hand, the spatial and temporal variability in catchment features should be adequately represented in the model parameterization, while maintaining the model complexity in an acceptable level to take advantage of state-of-the-art calibration techniques. On the other hand, model complexity enhances uncertainty in adjusted model parameter values, therefore increasing uncertainty in the water routing across the watershed. This is critical for water quality applications, where not only streamflow, but also a reliable estimation of the surface versus subsurface contributions to the runoff is needed. In this study, we show how a regularized inversion procedure combined with a multiobjective function calibration strategy successfully solves the parameterization of a complex application of a water quality-oriented hydrological model. The final value of several optimized parameters showed significant and consistentdifferences across geological and landscape features. Although the number of optimized parameters was significantly increased by the spatial and temporal discretization of adjustable parameters, the uncertainty in water routing results remained at reasonable values. In addition, a stepwise numerical analysis showed that the effects on calibration performance due to inclusion of different data types in the objective function could be inextricably linked. Thus caution should be taken when adding or removing data from an aggregated objective function.

Naevus du repli semi-lunaire de la conjonctive de découverte tardive [Conjunctival semi-lunar fold nevus of presumed late onset].

Conjunctival nevi are lesions observed commonly occasionnally and can be found later in a life. Such a lesion was observed on the conjunctival semilunar fold in a 76-year-old female patient. It was oval in shape, pigmented, and was mainly visible when the affected eye was in abduction. As no accurate anterior clinical history was available, surgical resection was performed. Hitopathological study disclosed a subepithelial nevus, which was completely resected. Regular follow-up was recommended, as is usual after surgical resection of a conjunctival pigmented tumor.

Supervised injection services : what has been demonstrated? A systematic literature review

BACKGROUND: Supervised injection services (SISs) have been developed to promote safer drug injection practices, enhance health-related behaviors among people who inject drugs (PWID), and connect PWID with external health and social services. Nevertheless, SISs have also been accused of fostering drug use and drug trafficking. AIMS: To systematically collect and synthesize the currently available evidence regarding SIS-induced benefits and harm. METHODS: A systematic review was performed via the PubMed, Web of Science, and ScienceDirect databases using the keyword algorithm [("supervised" or "safer") and ("injection" or "injecting" or "shooting" or "consumption") and ("facility" or "facilities" or "room" or "gallery" or "centre" or "site")]. RESULTS: Seventy-five relevant articles were found. All studies converged to find that SISs were efficacious in attracting the most marginalized PWID, promoting safer injection conditions, enhancing access to primary health care, and reducing the overdose frequency. SISs were not found to increase drug injecting, drug trafficking or crime in the surrounding environments. SISs were found to be associated with reduced levels of public drug injections and dropped syringes. Of the articles, 85% originated from Vancouver or Sydney. CONCLUSION: SISs have largely fulfilled their initial objectives without enhancing drug use or drug trafficking. Almost all of the studies found in this review were performed in Canada or Australia, whereas the majority of SISs are located in Europe. The implementation of new SISs in places with high rates of injection drug use and associated harms appears to be supported by evidence.

Astrocytic αVβ3 integrin inhibits neurite outgrowth and promotes retraction of neuronal processes by clustering Thy-1.

Thy-1 is a membrane glycoprotein suggested to stabilize or inhibit growth of neuronal processes. However, its precise function has remained obscure, because its endogenous ligand is unknown. We previously showed that Thy-1 binds directly to α(V)β(3) integrin in trans eliciting responses in astrocytes. Nonetheless, whether α(V)β(3) integrin might also serve as a Thy-1-ligand triggering a neuronal response has not been explored. Thus, utilizing primary neurons and a neuron-derived cell line CAD, Thy-1-mediated effects of α(V)β(3) integrin on growth and retraction of neuronal processes were tested. In astrocyte-neuron co-cultures, endogenous α(V)β(3) integrin restricted neurite outgrowth. Likewise, α(V)β(3)-Fc was sufficient to suppress neurite extension in Thy-1(+), but not in Thy-1(-) CAD cells. In differentiating primary neurons exposed to α(V)β(3)-Fc, fewer and shorter dendrites were detected. This effect was abolished by cleavage of Thy-1 from the neuronal surface using phosphoinositide-specific phospholipase C (PI-PLC). Moreover, α(V)β(3)-Fc also induced retraction of already extended Thy-1(+)-axon-like neurites in differentiated CAD cells as well as of axonal terminals in differentiated primary neurons. Axonal retraction occurred when redistribution and clustering of Thy-1 molecules in the plasma membrane was induced by α(V)β(3) integrin. Binding of α(V)β(3)-Fc was detected in Thy-1 clusters during axon retraction of primary neurons. Moreover, α(V)β(3)-Fc-induced Thy-1 clustering correlated in time and space with redistribution and inactivation of Src kinase. Thus, our data indicates that α(V)β(3) integrin is a ligand for Thy-1 that upon binding not only restricts the growth of neurites, but also induces retraction of already existing processes by inducing Thy-1 clustering. We propose that these events participate in bi-directional astrocyte-neuron communication relevant to axonal repair after neuronal damage.

Avaliação de um modelo físico de produção de bovinos no semi-árido integrando caatinga, capim-buffel e leucena.: I. fase de cria

O estudo teve como objetivo avaliar um modelo físico do sistema de cria de bovinos para as áreas mais secas do Semi-Árido brasileiro, associando a pastagem natural com pastos cultivados tolerantes à seca. A caatinga foi pastejada no período verde, e o capim buffel no restante do ano. A suplementação com leucena ocorreu no período seco, sob pastejo direto e feno no cocho. Parâmetros de desempenho, monitorados de novembro de 1991 a outubro de 1995, mostraram uma taxa média de parição da ordem de 72,8% ao ano e taxas de mortalidade praticamente nulas. O peso vivo médio dos bezerros aos 205 dias de idade foi de 153,4 kg e a produ- ção de bezerros desmamados foi de 109,5 kg/matriz exposta/ano e de 25,5 kg/hectare/ano. Os resultados são considerados bastante expressivos, considerando-se que mais da metade da área era ocupada com pastagens nativas de caatinga e que a oferta de leucena foi limitada pela ocorrência de uma forte estiagem.

Uses of also in oral semi-informal German

In recent grammars and dictionaries also (`therefore, so, well¿) continues to be preferably presented as an adverb with a conclusive-consecutive connective function that essentially corresponds to its use in formal written German. Its function as a modal particle is documented, however, since the beginnings of what is known as Partikelforschung, though not all its uses have been systematically investigated contrasting oral and written German, either in mode or concept. In this article we analyse the uses of also in semi-informal oral interactions on the basis of empirical data (from a subsample of the VARCOM corpus). Specifically, we will analyse the presence and frequency of also at the beginning of a sentence or sequence, the functions it serves as a logical-semantic connector or discourse and interaction marker and the interrelations between these functions, in order to contrast these results with the description of also provided by current reference works.

SUPLEMENTAÇÃO ALIMENTAR DE CABRAS ANGLO-NUBIANAS NA ÉPOCA CHUVOSA, NA REGIÃO SEMI-ÁRIDA DO NORDESTE BRASILEIRO

Foi realizado um experimento para avaliar a influência da suplementação alimentar na produção de leite de cabras da raça Anglo-nubiana, durante a época chuvosa, na região semi-árida do Nordeste brasileiro. Vinte e quatro cabras foram mantidas em pastagem nativa rebaixada, distribuídas ao acaso, por tipo de parto e produção de leite, em três níveis de suplementação: N0 - sem suplementação, N1 -150 g de concentrado/animal/dia e N2 - 300 g de concentrado/animal/dia. Não houve diferença significativa (p>0,05) entre os níveis de suplementação. A produção de leite foi descrita por uma regressão múltipla em função do período de lactação em semana. A interação semana x níveis de suplementação foi significativa (p<0,01). Os resultados revelam que a suplementação com concentrado, na dose de 300 g/animal/dia, foi a melhor opção para se produzir leite com cabras Anglo-nubianas, durante a época chuvosa.

Efeito da calagem e inoculação de sabiá em solo da mata úmida e do semi-árido de Pernambuco

Realizou-se um experimento, usando solo álico da Zona da Mata e do Semi-Árido do Estado de Pernambuco, com o objetivo de avaliar o efeito da calagem (0, 3 e 6 t ha-1) e da inoculação das estirpes de rizóbio: NFB 539, NFB 577 e NFB 578, previamente selecionadas para sabiá (Mimosa caesalpiniaefolia) em meio de cultura ácido. Adicionaram-se tratamentos sem inoculação, sem e com adição de N mineral (100 kg ha-1), para fins comparativos. O experimento seguiu o esquema fatorial 2x3x5, no delineamento em blocos ao acaso, com quatro repetições. As plantas foram colhidas 110 dias após a emergência (DAE). A inoculação de rizóbio, em todos os níveis de calagem, mostrou efeito significativo nos parâmetros avaliados. No nível de 3 t ha-1 não houve efeito da calagem, porém a adição de 6 t ha-1 de calcário reduziu o peso de matéria seca, o N total na parte aérea, a nodulação e a atividade da nitrogenase. O baixo pH e Al3+ do solo não prejudicou a fixação do N2, e o crescimento das plantas que receberam o inóculo. Portanto, torna-se desnecessária a calagem no cultivo de sabiá em solos ácidos, quando usadas estirpes selecionadas visando à resistência a acidez.

Efeito de inoculante ectomicorrízico produzido por fermentação semi-sólida sobre o crescimento de Eucalyptus dunnii Maiden

O crescimento de Eucalyptus é favorecido por fungos ectomicorrízicos (FECM). De ocorrência comum nas plantações florestais, os FECM diferem, entretanto, em infectividade e eficiência. O objetivo deste estudo foi testar a compatibilidade e a eficiência do inoculante ectomicorrízico produzido por fermentação semi-sólida, em relação a Eucalyptus dunnii Maiden. Neste sentido, o inoculante de Pisolithus sp., produzido por fermentação semi-sólida em vermiculita-solução nutritiva, foi aplicado ao substrato turfa-vermiculita (30:70, v/v) nas concentrações de 0, 1, 3, 5 e 10% (v/v) que foi semeado com E. dunnii, e mantido em casa de vegetação. A inoculação proporcionou colonização radicular e aumento da matéria seca da parte aérea e do conteúdo de P em razão da concentração do inoculante. O peso da matéria seca das plantas com 1% de inoculante não diferiu do peso das testemunhas. A 3% e 5%, a matéria seca aumentou 62%. O maior valor foi obtido com 10% de inoculante, onde o peso das plantas foi 73% superior ao das testemunhas, e o teor de P, 130%. Resultados semelhantes foram obtidos em relação à raiz. Altura e diâmetro foram positivamente afetados pela inoculação, mas não a relação raiz/parte aérea. As doses mais eficientes de inoculante foram aquelas de 3% ou superiores. Este método de produção deverá ser testado com outros FECM, e sua eficiência, avaliada em viveiro e no campo.

Large-scale clustering through functional embedding

We present a new framework for large-scale data clustering. The main idea is to modify functional dimensionality reduction techniques to directly optimize over discrete labels using stochastic gradient descent. Compared to methods like spectral clustering our approach solves a single optimization problem, rather than an ad-hoc two-stage optimization approach, does not require a matrix inversion, can easily encode prior knowledge in the set of implementable functions, and does not have an ?out-of-sample? problem. Experimental results on both artificial and real-world datasets show the usefulness of our approach.