Implementing statistical learning methods through Bayesian networks (Part 2): bayesian evaluations for results of black toner analyses in forensic document examination

This paper presents and discusses the use of Bayesian procedures - introduced through the use of Bayesian networks in Part I of this series of papers - for 'learning' probabilities from data. The discussion will relate to a set of real data on characteristics of black toners commonly used in printing and copying devices. Particular attention is drawn to the incorporation of the proposed procedures as an integral part in probabilistic inference schemes (notably in the form of Bayesian networks) that are intended to address uncertainties related to particular propositions of interest (e.g., whether or not a sample originates from a particular source). The conceptual tenets of the proposed methodologies are presented along with aspects of their practical implementation using currently available Bayesian network software.

Constructing brain functional networks from EEG: partial and unpartial correlations.

We consider electroencephalograms (EEGs) of healthy individuals and compare the properties of the brain functional networks found through two methods: unpartialized and partialized cross-correlations. The networks obtained by partial correlations are fundamentally different from those constructed through unpartial correlations in terms of graph metrics. In particular, they have completely different connection efficiency, clustering coefficient, assortativity, degree variability, and synchronization properties. Unpartial correlations are simple to compute and they can be easily applied to large-scale systems, yet they cannot prevent the prediction of non-direct edges. In contrast, partial correlations, which are often expensive to compute, reduce predicting such edges. We suggest combining these alternative methods in order to have complementary information on brain functional networks.

Connectivity differences in brain networks.

The scenario considered here is one where brain connectivity is represented as a network and an experimenter wishes to assess the evidence for an experimental effect at each of the typically thousands of connections comprising the network. To do this, a univariate model is independently fitted to each connection. It would be unwise to declare significance based on an uncorrected threshold of α=0.05, since the expected number of false positives for a network comprising N=90 nodes and N(N-1)/2=4005 connections would be 200. Control of Type I errors over all connections is therefore necessary. The network-based statistic (NBS) and spatial pairwise clustering (SPC) are two distinct methods that have been used to control family-wise errors when assessing the evidence for an experimental effect with mass univariate testing. The basic principle of the NBS and SPC is the same as supra-threshold voxel clustering. Unlike voxel clustering, where the definition of a voxel cluster is unambiguous, 'clusters' formed among supra-threshold connections can be defined in different ways. The NBS defines clusters using the graph theoretical concept of connected components. SPC on the other hand uses a more stringent pairwise clustering concept. The purpose of this article is to compare the pros and cons of the NBS and SPC, provide some guidelines on their practical use and demonstrate their utility using a case study involving neuroimaging data.

Seismic attenuation and velocity dispersion in heterogeneous partially saturated porous rocks

Using a numerical approach, we explore wave-induced fluid flow effects in partially saturated porous rocks in which the gas-water saturation patterns are governed by mesoscopic heterogeneities associated with the dry frame properties. The link between the dry frame properties and the gas saturation is defined by the assumption of capillary pressure equilibrium, which in the presence of heterogeneity implies that neighbouring regions can exhibit different levels of saturation. To determine the equivalent attenuation and phase velocity of the synthetic rock samples considered in this study, we apply a numerical upscaling procedure, which permits to take into account mesoscopic heterogeneities associated with the dry frame properties as well as spatially continuous variations of the pore fluid properties. The multiscale nature of the fluid saturation is taken into account by locally computing the physical properties of an effective fluid, which are then used for the larger-scale simulations. We consider two sets of numerical experiments to analyse such effects in heterogeneous partially saturated porous media, where the saturation field is determined by variations in porosity and clay content, respectively. In both cases we also evaluate the seismic responses of corresponding binary, patchy-type saturation patterns. Our results indicate that significant attenuation and modest velocity dispersion effects take place in this kind of media for both binary patchy-type and spatially continuous gas saturation patterns and in particular in the presence of relatively small amounts of gas. The numerical experiments also show that the nature of the gas distribution patterns is a critical parameter controlling the seismic responses of these environments, since attenuation and velocity dispersion effects are much more significant and occur over a broader saturation range for binary patchy-type gas-water distributions. This analysis therefore suggests that the physical mechanisms governing partial saturation should be accounted for when analysing seismic data in a poroelastic framework. In this context, heterogeneities associated with the dry frame properties, which do not play important roles in wave-induced fluid flow processes per se, should be taken into account since they may determine the kind of gas distribution pattern taking place in the porous rock.

Bgee: integrating and comparing heterogeneous transcriptome data among species

Gene expression patterns are a key feature in understanding gene function, notably in development. Comparing gene expression patterns between animals is a major step in the study of gene function as well as of animal evolution. It also provides a link between genes and phenotypes. Thus we have developed Bgee, a database designed to compare expression patterns between animals, by implementing ontologies describing anatomies and developmental stages of species, and then designing homology relationships between anatomies and comparison criteria between developmental stages. To define homology relationships between anatomical features we have developed the software Homolonto, which uses a modified ontology alignment approach to propose homology relationships between ontologies. Bgee then uses these aligned ontologies, onto which heterogeneous expression data types are mapped. These already include microarrays and ESTs.

Monitoring of illicit pill distribution networks using an image collection exploration framework

This paper proposes a novel approach for the analysis of illicit tablets based on their visual characteristics. In particular, the paper concentrates on the problem of ecstasy pill seizure profiling and monitoring. The presented method extracts the visual information from pill images and builds a representation of it, i.e. it builds a pill profile based on the pill visual appearance. Different visual features are used to build different image similarity measures, which are the basis for a pill monitoring strategy based on both discriminative and clustering models. The discriminative model permits to infer whether two pills come from the same seizure, while the clustering models groups of pills that share similar visual characteristics. The resulting clustering structure allows to perform a visual identification of the relationships between different seizures. The proposed approach was evaluated using a data set of 621 Ecstasy pill pictures. The results demonstrate that this is a feasible and cost effective method for performing pill profiling and monitoring.

Kinship in entrepreneur networks: Performance effects of resource assembly in Africa

We examine the relationship between structural social capital, resource assembly, and firm performance of entrepreneurs in Africa. We posit that social capital primarily composed of kinship or family ties helps the entrepreneur to raise resources, but it does so at a cost. Using data drawn from small firms in Kampala, Uganda, we explore how shared identity among the entrepreneur's social network moderates this relationship. A large network contributed a higher quantity of resources raised, but at a higher cost when shared identity was high. We discuss the implications of these findings for the role of family ties and social capital in resource assembly, with an emphasis on developing economies.

The challenge of modeling nuclear receptor regulatory networks in mammalian cells.

Nuclear receptors are a major component of signal transduction in animals. They mediate the regulatory activities of many hormones, nutrients and metabolites on the homeostasis and physiology of cells and tissues. It is of high interest to model the corresponding regulatory networks. While molecular and cell biology studies of individual promoters have provided important mechanistic insight, a more complex picture is emerging from genome-wide studies. The regulatory circuitry of nuclear receptor regulated gene expression networks, and their response to cellular signaling, appear highly dynamic, and involve long as well as short range chromatin interactions. We review how progress in understanding the kinetics and regulation of cofactor recruitment, and the development of new genomic methods, provide opportunities but also a major challenge for modeling nuclear receptor mediated regulatory networks.

Bayesian networks for evaluating forensic DNA profiling evidence: a review and guide to literature

Almost 30 years ago, Bayesian networks (BNs) were developed in the field of artificial intelligence as a framework that should assist researchers and practitioners in applying the theory of probability to inference problems of more substantive size and, thus, to more realistic and practical problems. Since the late 1980s, Bayesian networks have also attracted researchers in forensic science and this tendency has considerably intensified throughout the last decade. This review article provides an overview of the scientific literature that describes research on Bayesian networks as a tool that can be used to study, develop and implement probabilistic procedures for evaluating the probative value of particular items of scientific evidence in forensic science. Primary attention is drawn here to evaluative issues that pertain to forensic DNA profiling evidence because this is one of the main categories of evidence whose assessment has been studied through Bayesian networks. The scope of topics is large and includes almost any aspect that relates to forensic DNA profiling. Typical examples are inference of source (or, 'criminal identification'), relatedness testing, database searching and special trace evidence evaluation (such as mixed DNA stains or stains with low quantities of DNA). The perspective of the review presented here is not exclusively restricted to DNA evidence, but also includes relevant references and discussion on both, the concept of Bayesian networks as well as its general usage in legal sciences as one among several different graphical approaches to evidence evaluation.

Presence of an oligodendroglioma-like component in newly diagnosed glioblastoma identifies a pathogenetically heterogeneous subgroup and lacks prognostic value: central pathology review of the EORTC_26981/NCIC_CE.3 trial.

Glioblastoma (GBM) is a morphologically heterogeneous tumor type with a median survival of only 15 months in clinical trial populations. However, survival varies greatly among patients. As part of a central pathology review, we addressed the question if patients with GBM displaying distinct morphologic features respond differently to combined chemo-radiotherapy with temozolomide. Morphologic features were systematically recorded for 360 cases with particular focus on the presence of an oligodendroglioma-like component and respective correlations with outcome and relevant molecular markers. GBM with an oligodendroglioma-like component (GBM-O) represented 15% of all confirmed GBM (52/339) and was not associated with a more favorable outcome. GBM-O encompassed a pathogenetically heterogeneous group, significantly enriched for IDH1 mutations (19 vs. 3%, p = 0.003) and EGFR amplifications (71 vs. 48%, p = 0.04) compared with other GBM, while co-deletion of 1p/19q was found in only one case and the MGMT methylation frequency was alike (47 vs. 46%). Expression profiles classified most of the GBM-O into two subtypes, 36% (5/14 evaluable) as proneural and 43% as classical GBM. The detection of pseudo-palisading necrosis (PPN) was associated with benefit from chemotherapy (p = 0.0002), while no such effect was present in the absence of PPN (p = 0.86). In the adjusted interaction model including clinical prognostic factors and MGMT status, PPN was borderline nonsignificant (p = 0.063). Taken together, recognition of an oligodendroglioma-like component in an otherwise classic GBM identifies a pathogenetically mixed group without prognostic significance. However, the presence of PPN may indicate biological features of clinical relevance for further improvement of therapy.

Character Networks and Centrality

A character network represents relations between characters from a text; the relations are based on text proximity, shared scenes/events, quoted speech, etc. Our project sketches a theoretical framework for character network analysis, bringing together narratology, both close and distant reading approaches, and social network analysis. It is in line with recent attempts to automatise the extraction of literary social networks (Elson, 2012; Sack, 2013) and other studies stressing the importance of character- systems (Woloch, 2003; Moretti, 2011). The method we use to build the network is direct and simple. First, we extract co-occurrences from a book index, without the need for text analysis. We then describe the narrative roles of the characters, which we deduce from their respective positions in the network, i.e. the discourse. As a case study, we use the autobiographical novel Les Confessions by Jean-Jacques Rousseau. We start by identifying co-occurrences of characters in the book index of our edition (Slatkine, 2012). Subsequently, we compute four types of centrality: degree, closeness, betweenness, eigenvector. We then use these measures to propose a typology of narrative roles for the characters. We show that the two parts of Les Confessions, written years apart, are structured around mirroring central figures that bear similar centrality scores. The first part revolves around the mentor of Rousseau; a figure of openness. The second part centres on a group of schemers, depicting a period of deep paranoia. We also highlight characters with intermediary roles: they provide narrative links between the societies in the life of the author. The method we detail in this complete case study of character network analysis can be applied to any work documented by an index. Un réseau de personnages modélise les relations entre les personnages d'un récit : les relations sont basées sur une forme de proximité dans le texte, l'apparition commune dans des événements, des citations dans des dialogues, etc. Notre travail propose un cadre théorique pour l'analyse des réseaux de personnages, rassemblant narratologie, close et distant reading, et analyse des réseaux sociaux. Ce travail prolonge les tentatives récentes d'automatisation de l'extraction de réseaux sociaux tirés de la littérature (Elson, 2012; Sack, 2013), ainsi que les études portant sur l'importance des systèmes de personnages (Woloch, 2003; Moretti, 2011). La méthode que nous utilisons pour construire le réseau est directe et simple. Nous extrayons les co-occurrences d'un index sans avoir recours à l'analyse textuelle. Nous décrivons les rôles narratifs des personnages en les déduisant de leurs positions relatives dans le réseau, donc du discours. Comme étude de cas, nous avons choisi le roman autobiographique Les Confessions, de Jean- Jacques Rousseau. Nous déduisons les co-occurrences entre personnages de l'index présent dans l'édition Slatkine (Rousseau et al., 2012). Sur le réseau obtenu, nous calculons quatre types de centralité : le degré, la proximité, l'intermédiarité et la centralité par vecteur propre. Nous utilisons ces mesures pour proposer une typologie des rôles narratifs des personnages. Nous montrons que les deux parties des Confessions, écrites à deux époques différentes, sont structurées autour de deux figures centrales, qui obtiennent des mesures de centralité similaires. La première partie est construite autour du mentor de Rousseau, qui a symbolisé une grande ouverture. La seconde partie se focalise sur un groupe de comploteurs, et retrace une période marquée par la paranoïa chez l'auteur. Nous mettons également en évidence des personnages jouant des rôles intermédiaires, et de fait procurant un lien narratif entre les différentes sociétés couvrant la vie de l'auteur. La méthode d'analyse des réseaux de personnages que nous décrivons peut être appliquée à tout texte de fiction comportant un index.

Networks of Coordination : Swiss Business Associations as an Intermediary between Business, Politics and Administration during the 20th Century

Until the 1990's, Switzerland could be classified as either a corporatist, cooperative or coordinated market economy where non-market mechanisms of coordination among economic and political actors were very important. In this respect, Business Interest Associations (BIAs) played a key role. The aim of this paper is to look at the historical evolution of the five main peak Swiss BIAs through network analysis for five assorted dates during the 20th century (1910, 1937, 1957, 1980 and 2000) while relying on a database that includes more than 12,000 people. First, we examine the logic of membership in these associations, which allows us to analyze their position and function within the network of the Swiss economic elite. Until the 1980's, BIAs took part in the emergence and consolidation of a closely meshed national network, which declined during the two last decades of the 20th century. Second, we investigate the logic of influence of these associations by looking at the links they maintained with the political and administrative worlds through their links to the political parties and Parliament, and to the administration via the extra-parliamentary commissions (corporatist bodies). In both cases, the recent dynamic of globalization called into question the traditional role of BIAs.