Graph-based Vision-Language Model for Multi-Modal Retrieval of Fashion Products

Artificial Intelligence is reshaping the field of fashion industry in different ways. E-commerce retailers exploit their data through AI to enhance their search engines, make outfit suggestions and forecast the success of a specific fashion product. However, it is a challenging endeavour as the data they possess is huge, complex and multi-modal. The most common way to search for fashion products online is by matching keywords with phrases in the product's description which are often cluttered, inadequate and differ across collections and sellers. A customer may also browse an online store's taxonomy, although this is time-consuming and doesn't guarantee relevant items. With the advent of Deep Learning architectures, particularly Vision-Language models, ad-hoc solutions have been proposed to model both the product image and description to solve this problems. However, the suggested solutions do not exploit effectively the semantic or syntactic information of these modalities, and the unique qualities and relations of clothing items. In this work of thesis, a novel approach is proposed to address this issues, which aims to model and process images and text descriptions as graphs in order to exploit the relations inside and between each modality and employs specific techniques to extract syntactic and semantic information. The results obtained show promising performances on different tasks when compared to the present state-of-the-art deep learning architectures.

Tecniche innovative microbiome-based per il recupero della vongola comune nelle aree ad alta mortalità nell’area FLAG dell’Emilia Romagna

Negli ultimi decenni sono state registrate preoccupati fenomeni di mortalità della vongola Chamelea gallina, in particolare nell’area costiera emiliano-romagnola e di cui non sono ancora state chiarite le cause. Il presente studio si è occupato di caratterizzare la comunità microbica associata alla vongola nella ghiandola digestiva, utilizzando il sequenziamento della regione ipervariabile V3-V4 del gene rRNA 16S, al fine di individuare fenomeni di disbiosi in aree ad elevata mortalità. Sono state quindi esplorate le variazioni stagionali (da luglio a novembre) nella struttura del microbiota della vongola e nell'ecosistema microbico dell'acqua di mare circostante, in quattro siti scelti ad hoc, secondo un gradiente di incidenza storica di mortalità, da Nord a Sud, tra le aree di Ravenna e Rimini. Lo stato di salute della vongola e del suo microbiota associato sono stati esplorati tramite, rispettivamente, l’indice di condizione e lo studio mediante NGS della composizione dell’ecosistema microbico intestinale. I nostri dati, sebbene preliminari, dimostrano come tra le aree Nord e Sud ci sia un comportamento differente e reciproco relativamente all’andamento stagionale dei valori di diversità interna (alfa) al microbiota della vongola, che si riduce dall’estate all’autunno nelle aree Nord (Ravenna e Lido di Savio), mentre aumenta - nello stesso periodo di tempo - nelle aree Sud (Rimini e Cesenatico). A conferma dei dati di alfa diversità, l’analisi mediante PCoA delle variazione del microbiota della vongola tra i quattro siti di indagine stratificate per stagione, dimostrano profonde differenze tra i due estremi nord-sud. In particolare, l’analisi integrata dei dati storici di produttività, indice di condizione e dinamica del microbiota della g.d. ci ha consentito di discriminare cinque famiglie microbiche come potenziali Growth Promoting Bacteria, poiché associate ad un picco di indice di condizione che si registra nelle aree a bassa mortalità, nel mese di settembre.

Advancing Bag-of-visual-words Representations For Lesion Classification In Retinal Images.

Diabetic Retinopathy (DR) is a complication of diabetes that can lead to blindness if not readily discovered. Automated screening algorithms have the potential to improve identification of patients who need further medical attention. However, the identification of lesions must be accurate to be useful for clinical application. The bag-of-visual-words (BoVW) algorithm employs a maximum-margin classifier in a flexible framework that is able to detect the most common DR-related lesions such as microaneurysms, cotton-wool spots and hard exudates. BoVW allows to bypass the need for pre- and post-processing of the retinographic images, as well as the need of specific ad hoc techniques for identification of each type of lesion. An extensive evaluation of the BoVW model, using three large retinograph datasets (DR1, DR2 and Messidor) with different resolution and collected by different healthcare personnel, was performed. The results demonstrate that the BoVW classification approach can identify different lesions within an image without having to utilize different algorithms for each lesion reducing processing time and providing a more flexible diagnostic system. Our BoVW scheme is based on sparse low-level feature detection with a Speeded-Up Robust Features (SURF) local descriptor, and mid-level features based on semi-soft coding with max pooling. The best BoVW representation for retinal image classification was an area under the receiver operating characteristic curve (AUC-ROC) of 97.8% (exudates) and 93.5% (red lesions), applying a cross-dataset validation protocol. To assess the accuracy for detecting cases that require referral within one year, the sparse extraction technique associated with semi-soft coding and max pooling obtained an AUC of 94.2 ± 2.0%, outperforming current methods. Those results indicate that, for retinal image classification tasks in clinical practice, BoVW is equal and, in some instances, surpasses results obtained using dense detection (widely believed to be the best choice in many vision problems) for the low-level descriptors.

Análise do jogo de goalball : modelação e interpretação dos padrões de jogo da Paralimpíada de Pequim 2008

Universidade Estadual de Campinas . Faculdade de Educação Física

Hubble parameter reconstruction from a principal component analysis: minimizing the bias

Aims. A model-independent reconstruction of the cosmic expansion rate is essential to a robust analysis of cosmological observations. Our goal is to demonstrate that current data are able to provide reasonable constraints on the behavior of the Hubble parameter with redshift, independently of any cosmological model or underlying gravity theory. Methods. Using type Ia supernova data, we show that it is possible to analytically calculate the Fisher matrix components in a Hubble parameter analysis without assumptions about the energy content of the Universe. We used a principal component analysis to reconstruct the Hubble parameter as a linear combination of the Fisher matrix eigenvectors (principal components). To suppress the bias introduced by the high redshift behavior of the components, we considered the value of the Hubble parameter at high redshift as a free parameter. We first tested our procedure using a mock sample of type Ia supernova observations, we then applied it to the real data compiled by the Sloan Digital Sky Survey (SDSS) group. Results. In the mock sample analysis, we demonstrate that it is possible to drastically suppress the bias introduced by the high redshift behavior of the principal components. Applying our procedure to the real data, we show that it allows us to determine the behavior of the Hubble parameter with reasonable uncertainty, without introducing any ad-hoc parameterizations. Beyond that, our reconstruction agrees with completely independent measurements of the Hubble parameter obtained from red-envelope galaxies.

Thermodynamic modeling of pyroelectric and piezoelectric properties of cellular polymers

A thermodynamic approach is presented to model devices manufactured with cellular polymers. They are heterogeneous nonpolar space-charge electrets that exhibit much higher piezoelectricity than the well-known ferroelectric polymers. Their pyroelectric and piezoelectric properties are characterized by adequate coefficients which quantify the performance of devices manufactured with those materials. The method presented in this contribution to calculate those coefficients is exact and consistent avoiding ad hoc simplifications introduced in other approaches. The results obtained by this method allow drawing conclusions regarding device optimization.

Spatio-temporal modeling of agricultural yield data with an application to pricing crop insurance contracts

This article presents a statistical model of agricultural yield data based on a set of hierarchical Bayesian models that allows joint modeling of temporal and spatial autocorrelation. This method captures a comprehensive range of the various uncertainties involved in predicting crop insurance premium rates as opposed to the more traditional ad hoc, two-stage methods that are typically based on independent estimation and prediction. A panel data set of county-average yield data was analyzed for 290 counties in the State of Parana (Brazil) for the period of 1990 through 2002. Posterior predictive criteria are used to evaluate different model specifications. This article provides substantial improvements in the statistical and actuarial methods often applied to the calculation of insurance premium rates. These improvements are especially relevant to situations where data are limited.

Popper and his Commentators on the Discovery of Neptune: a close shave for the Law of Gravitation?

Knowledge of residual perturbations in the orbit of Uranus in the early 1840s did not lead to the refutation of Newton's law of gravitation but instead to the discovery of Neptune in 1846. Karl Popper asserts that this case is atypical of science and that the law of gravitation was at least prima facie falsified by these perturbations. I argue that these assertions are the product of a false, a priori methodological position I call, 'Weak Popperian Falsificationism' (WPF). Further, on the evidence the law was not prima facie false and was not generally considered so by astronomers at the time. Many of Popper's commentators (Kuhn, Lakatos, Feyerabend and others) presuppose WPF and their views on this case and its implications for scientific rationality and method suffer from this same defect.

Allochronic taxa as an alternative model to explain circumantarctic disjunctions

Most biogeographical studies propose that southern temperate faunal disjunctions are either the result of vicariance of taxa originated in Gondwana or the result of transoceanic dispersal of taxa originated after the breakup of Gondwana. The aim of this paper is to show that this is a false dichotomy. Antarctica retained a mild climate until mid-Cenozoic and had lasting connections, notably with southern South America and Australia. Both taxa originally Gondwanan and taxa secondarily on Gondwanan areas were subjected to tectonic-induced vicariance, and there is no need to invoke ad hoc transoceanic dispersal, even for post-Gondwanan taxa. These different elements with circumantarctic distributions are here called `allochronic taxa` - taxa presently occupying the same area, but whose presence in that area does not belong to the same time period. This model allows accommodation of conflicting sources of evidence now available for many groups with circumantarctic distributions. The fact that the species from both layers are mixed up in the current biodiversity implies the need to use additional sources of evidence - such as biogeographical, palaeontological, geological and molecular - to discriminate which are the original Gondwanan and which are post-Gondwanan elements in austral landmasses.

Labour flexibility and gender in the service sector: A study of employment practices in law and accounting firms

In seeking to increase the flexibility of their use of employee time, employers can pursue strategies based on the employment of casual and part-time workers (numerical flexibility) or strategies based on ad hoc variation of the working hours of permanent employees (working time flexibility). Patterns of flexibility strategies and their implications are examined in the context of a highly feminised sector of work-clerical and administrative employment in law and accounting firms. We consider whether, as is often assumed, working time flexibility strategies are generally better for employees because they avoid the substitution of core, high quality jobs with the peripheral, relatively insecure employment often associated with casualisation. Analysing data drawn from a survey of law and accounting firms, we argue that there are three distinct flexibility strategies adopted by employers, and that the choice of strategy is influenced by the size of the firm and the extent of feminisation. The quality of employment conditions associated with each strategy is investigated through an analysis of the determinants of training provision for clerical and administrative workers. Rather than an expected simple linear relationship between increasing casualisation and decreasing training provision, we find that firm size and feminisation are implicated. Larger firms that tend to employ at least some men and use a combination of working time and numerical flexibility strategies tend to provide more training than the small, more fully feminised firms that tend to opt for either casualisation or working time flexibility strategies. This suggests that, from an employee perspective, working time flexibility may not be as benevolent as is often thought.

Discrete and continuous models for dry masonry columns

The dynamic response of dry masonry columns can be approximated with finite-difference equations. Continuum models follow by replacing the difference quotients of the discrete model by corresponding differential expressions. The mathematically simplest of these models is a one-dimensional Cosserat theory. Within the presented homogenization context, the Cosserat theory is obtained by making ad hoc assumptions regarding the relative importance of certain terms in the differential expansions. The quality of approximation of the various theories is tested by comparison of the dispersion relations for bending waves with the dispersion relation of the discrete theory. All theories coincide with differences of less than 1% for wave-length-block-height (L/h) ratios bigger than 2 pi. The theory based on systematic differential approximation remains accurate up to L/h = 3 and then diverges rapidly. The Cosserat model becomes increasingly inaccurate for L/h < 2 pi. However, in contrast to the systematic approximation, the wave speed remains finite. In conclusion, considering its relative simplicity, the Cosserat model appears to be the natural starting point for the development of continuum models for blocky structures.

Phylogenetic relationships and uncertain taxonomy of Pedomicrobium species

The phylogenetic relationships among the species of the genus Pedomicrobium were studied by comparing their 16S rRNA sequences, The Pedomicrobium species form a coherent phylogenetic cluster within the genera of the hyphal budding bacteria in the alpha-Proteobacteria. The sequences of two strains of Pedomicrobium australicum were obtained from DNAs extracted from nonviable freeze-dried cells, which are the only source of material available, and were found to be almost identical (level of similarity, 99.9%), Overall, the Pedomicrobium species are closely related, with sequence similarities ranging from 96.2 to 99.9%, Pedomicrobium manganicum is phylogenetically the most distantly related species and exhibits the lowest similarity (96.2%) with Pedomicrobium americanum, Australian isolate Pedomicrobium sp, strain ACM 3067, P, americanum, and P, australicum are all very highly related, with similarities greater than 99%, Pedomicrobium sp, strain ACM 3067 is most closely related to P, australicum (level of similarity, 99.6%) and P, americanum (99.4%), These manganese-oxidizing species are more closely related to the iron-oxidizing species Pedomicrobium ferrugineum than to the other manganese-oxidizing species, P, manganicum. Taxonomic uncertainties resulting from the loss of the type culture of P, australicum are discussed.

Associativismo intergovernamental: experi??ncias brasileiras

O oitavo volume da s??rie Inova????o na Gest??o P??blica apresenta o resultado de uma pesquisa desenvolvida entre os anos de 2010 e 2011, na qual foram realizados onze estudos de caso sobre formas de arranjos associativos intergovernamentais existentes no Brasil, e contou com o apoio do Projeto Brasil Munic??pios. A expectativa com essa discuss??o ?? que se possam trazer ?? tona novos temas ou abordagens sobre associativismo, problematizar alguns importantes desafios, captar poss??veis funcionamentos das l??gicas de articula????o intergovernamental e possibilitar a reflex??o sobre alternativas para coordena????o e coopera????o intergovernamentais no Brasil. Dessa forma, o livro inicia-se por uma discuss??o te??rica com o objetivo de real??ar a import??ncia dos modelos de entrela??amento e coopera????o entre os n??veis de governo, sem que estes percam a autonomia e mesmo a capacidade de ???competir por agendas de pol??ticas???. Por outro lado, busca mostrar que formas de parceria e de atua????o em rede do governo para com a sociedade t??m se acoplado ao associativismo territorial. O federalismo ?? apresentado tamb??m sob uma perspectiva internacional, sobretudo fazendo um apanhado geral dos mecanismos de associativismo territorial em pa??ses federativos ou semifederativos, com destaque principal para os casos dos Estados Unidos, da Alemanha, da Espanha, do Canad?? e da Austr??lia, buscando ressaltar, principalmente, dos aspectos: o primeiro deles ?? apresentar a diversidade de possibilidades de coopera????o intergovernamental. O livro apresenta ainda uma an??lise do federalismo brasileiro contempor??neo, come??ando com um hist??rico resumido sobre a Federa????o brasileira, com foco no estudo dos processos de coordena????o e coopera????o intergovernamental existentes. Adicionalmente, apresenta-se um mapeamento das experi??ncias de associativismo territorial no Brasil, a partir de uma tipologia de arranjos que foram vinculados a dez diferentes l??gicas de articula????o intergovernamental: l??gica setorial; l??gica metropolitana; l??gica de desenvolvimento regional contra a desigualdade; l??gica de desenvolvimento regional baseada na identidade territorial; l??gica de atua????o regional da uni??o e dos governos estaduais; l??gica de atua????o microrregional; l??gica econ??mica de parceria p??blico-privado; l??gica social de parceria p??blico-privado; l??gica de associativismo de advocacy e coopera????o intergovernamental; e l??gica ad hoc de coopera????o intergovernamental. Por fim, apresenta-se uma s??ntese de onze estudos de caso sobre associativismo territorial no Brasil, com o intuito de demonstrar as poss??veis a????es que estariam ligadas as tipos de l??gica de articula????o intergovernamental, bem como propor a????es e incentivos nos plano institucional, da pol??ticas p??blicas e da governan??a territorial ampliada, a fim de buscar poss??veis solu????es para os problemas e dificuldades abordados por essas experi??ncias associativas

Análise nutricional e complementação alimentar de cesta básica derivada do consumo

INTRODUÇÃO: A "cesta básica Dieese e Procon", originada de uma Pesquisa de Padrão de Vida e Emprego no Município de São Paulo, tem sido empregada como parâmetro para o acompanhamento de preços. Seria desejável que um instrumento econômico, utilizado com esta finalidade, correspondesse também a uma nutrição efetivamente saudável. Assim, foram analisados os níveis de adequação dos itens alimentares da cesta básica em relação às necessidades nutricionais de uma família-referência paulistana, e propostas técnicas de complementação dietética para sanar possíveis deficiências ou desbalanceamento. METODOLOGIA: Utilizou-se o Censo Demográfico do IBGE, de 1991, para se determinar a família-referência; adotaram-se dois terços das "Recommended Dietary Allowances" como parâmetro de necessidades nutricionais; e utilizou-se das tabelas de composição centesimal para verificação do aporte de nutrientes da cesta básica. Efetuou-se a complementação alimentar por meio de três diferentes métodos: Ad Hoc (proposta direta e fixa), Programação Linear (proposta via computador e sazonal, com ênfase na minimização de custos), Híbrido (uma combinação dos dois anteriores). RESULTADOS: Foram encontrados valores insuficientes para as vitaminas A, C, B2 e B6 e para os minerais Ca, Mg, Fe, Zn, I e Se; o percentual de lipídios no total calórico mostrou-se elevado. Entre as técnicas de complementação, o método Híbrido pareceu assimilar, mais eficientemente, os baixos custos e os hábitos dietéticos. DISCUSSÃO: Inferiu-se uma eventual correspondência entre os problemas nutricionais detectados na relação de itens da cesta básica e a chamada transição alimentar que se processa nos países em desenvolvimento. Não parece aconselhável que os riscos epidemiológicos associados a essa alteração nos padrões dietéticos sejam incorporados num instrumento econômico que tenha por finalidade mensurar os preços de uma alimentação equilibrada.

Amostra mestra e geoprocessamento: tecnologias para inquéritos domiciliares

OBJETIVO: Reduzir as despesas e o tempo associados ao processo de amostragem de domicílios e mostrar a viabilidade de uso compartilhado dos cadastros de endereços entre vários inquéritos epidemiológicos, a partir da reciclagem de material produzido pela Pesquisa Nacional de Amostras de Domicílios (PNAD). MÉTODOS: Cadastros permanentes de endereços foram construídos e atualizados para 72 setores censitários, e retidos como unidades primárias da amostra mestra do município de São Paulo, SP. No período 1995-2000, sortearam-se três diferentes amostras pelo processo de conglomerados, em dois estágios. Operações de sorteio e mapeamento das amostras foram automatizadas por meio da aplicação de tecnologia de geoprocessamento. RESULTADOS: Cerca de 25.000 domicílios compõem o cadastro permanente de endereços da amostra mestra. A produção ágil e barata de cada amostra, aliada ao conhecimento acumulado sobre os perfis demográficos e topográficos dos setores censitários, pode ser considerada a principal contribuição dos resultados alcançados. Verificou-se que o sistema proposto contrasta com a seleção de amostras independentes para cada investigação "ad hoc", o que ainda é prática comum nas agências ou grupos que realizam inquéritos domiciliares. CONCLUSÕES: A reciclagem das listagens anuais de endereços da PNAD, apesar de restringir a abrangência de replicação da experiência aos municípios mais populosos do País, evita a repetição do processo de amostragem para cada inquérito; além disso, significa importante contribuição ao planejamento de investigações descritivas na área da saúde pública. Questões conceituais ou teóricas, decorrentes do procedimento operacional, foram facilmente superadas pela introdução de técnicas estatísticas compensatórias.