Decision boundary for discrete Bayesian network classifiers

Bayesian network classifiers are a powerful machine learning tool. In order to evaluate the expressive power of these models, we compute families of polynomials that sign-represent decision functions induced by Bayesian network classifiers. We prove that those families are linear combinations of products of Lagrange basis polynomials. In absence of V-structures in the predictor sub-graph, we are also able to prove that this family of polynomials does in- deed characterize the specific classifier considered. We then use this representation to bound the number of decision functions representable by Bayesian network classifiers with a given structure and we compare these bounds to the ones obtained using Vapnik-Chervonenkis dimension.

Decision boundary for discrete Bayesian network classifiers

Bayesian network classifiers are a powerful machine learning tool. In order to evaluate the expressive power of these models, we compute families of polynomials that sign-represent decision functions induced by Bayesian network classifiers. We prove that those families are linear combinations of products of Lagrange basis polynomials. In absence of V-structures in the predictor sub-graph, we are also able to prove that this family of polynomials does in- deed characterize the specific classifier considered. We then use this representation to bound the number of decision functions representable by Bayesian network classifiers with a given structure and we compare these bounds to the ones obtained using Vapnik-Chervonenkis dimension.

Decision boundary for discrete Bayesian network classifiers

Bayesian network classifiers are a powerful machine learning tool. In order to evaluate the expressive power of these models, we compute families of polynomials that sign-represent decision functions induced by Bayesian network classifiers. We prove that those families are linear combinations of products of Lagrange basis polynomials. In absence of V -structures in the predictor sub-graph, we are also able to prove that this family of polynomials does indeed characterize the specific classifier considered. We then use this representation to bound the number of decision functions representable by Bayesian network classifiers with a given structure.

From the outside in:consumer anti-choice and policy implications in the mobile gaming market

Significant growth in mobile media consumption has prompted a call to better understand the socio-cultural and policy dimensions of consumer choices. Contrary to industry and technology led analysis, this study argues that to guide consumer choice and innovation via regulatory policies requires an understanding of both ex-ante as well as in ex-post consumption conditions. This study examines mobile phone gaming to uncover how consumer anti-choice shapes decision-making as a framework for closely interrogating the ways in which policy concerns impact on consumers' behavior. Through eleven focus groups (n=62), the study empirically identifies voluntary, intentional, and positive consumer anti-choice behaviors all of which impact policy initiatives when consumers, both gamers and non-gamers, self-regulate their behaviors. Findings point to four types of policy implication: regulating the self-regulated, understanding anti-choice, boundary-setting and including the self-excluded. © 2012 Elsevier Ltd.

Truncated robust distance for clinical laboratory safety data monitoring and assessment.

Laboratory safety data are routinely collected in clinical studies for safety monitoring and assessment. We have developed a truncated robust multivariate outlier detection method for identifying subjects with clinically relevant abnormal laboratory measurements. The proposed method can be applied to historical clinical data to establish a multivariate decision boundary that can then be used for future clinical trial laboratory safety data monitoring and assessment. Simulations demonstrate that the proposed method has the ability to detect relevant outliers while automatically excluding irrelevant outliers. Two examples from actual clinical studies are used to illustrate the use of this method for identifying clinically relevant outliers.

A-optimality orthogonal forward regression algorithm using branch and bound

In this brief, we propose an orthogonal forward regression (OFR) algorithm based on the principles of the branch and bound (BB) and A-optimality experimental design. At each forward regression step, each candidate from a pool of candidate regressors, referred to as S, is evaluated in turn with three possible decisions: 1) one of these is selected and included into the model; 2) some of these remain in S for evaluation in the next forward regression step; and 3) the rest are permanently eliminated from S. Based on the BB principle in combination with an A-optimality composite cost function for model structure determination, a simple adaptive diagnostics test is proposed to determine the decision boundary between 2) and 3). As such the proposed algorithm can significantly reduce the computational cost in the A-optimality OFR algorithm. Numerical examples are used to demonstrate the effectiveness of the proposed algorithm.

A Robust, Fully Adaptive Hybrid Level-Set/Front-Tracking Method for Two-Phase Flows with an Accurate Surface Tension Computation

We present a variable time step, fully adaptive in space, hybrid method for the accurate simulation of incompressible two-phase flows in the presence of surface tension in two dimensions. The method is based on the hybrid level set/front-tracking approach proposed in [H. D. Ceniceros and A. M. Roma, J. Comput. Phys., 205, 391400, 2005]. Geometric, interfacial quantities are computed from front-tracking via the immersed-boundary setting while the signed distance (level set) function, which is evaluated fast and to machine precision, is used as a fluid indicator. The surface tension force is obtained by employing the mixed Eulerian/Lagrangian representation introduced in [S. Shin, S. I. Abdel-Khalik, V. Daru and D. Juric, J. Comput. Phys., 203, 493-516, 2005] whose success for greatly reducing parasitic currents has been demonstrated. The use of our accurate fluid indicator together with effective Lagrangian marker control enhance this parasitic current reduction by several orders of magnitude. To resolve accurately and efficiently sharp gradients and salient flow features we employ dynamic, adaptive mesh refinements. This spatial adaption is used in concert with a dynamic control of the distribution of the Lagrangian nodes along the fluid interface and a variable time step, linearly implicit time integration scheme. We present numerical examples designed to test the capabilities and performance of the proposed approach as well as three applications: the long-time evolution of a fluid interface undergoing Rayleigh-Taylor instability, an example of bubble ascending dynamics, and a drop impacting on a free interface whose dynamics we compare with both existing numerical and experimental data.

Financiamento para o subdesenvolvimento, o Brasil e o segundo consenso de Washington

While in the social and in the ethical realms the Cardoso administration was successful, its economic outcomes were frustrating. In this administration’s eight years the investment rate did not increase and income per capita growth lagged, while the public debt and the foreign debt increased substantially. This poor economic performance may be explained by three chained causes: a mistake in agenda setting, the adoption of the Second Washington Consensus, and the alienation of elites. The decision of setting high inflation as the major problem to be tackled instead of achieving equilibrium in foreign accounts represented a major macroeconomic mistake, which can be explained by the Second Washington Consensus. This consensus proposed in the 1990s that highly indebted countries should grow counting on foreign savings, although this is not the experience among OECD countries. The outcome was to evaluate the real, to increase artificially wages and consumption, so that instead of growth what we have been increased indebtedness. Why this flopped strategy was adopted? Rich countries’ interests are not difficult to guess. On the part of Brazil, the only explanation is Brazilian elites’ alienation in relation to the country’s national interest. As a final outcome, the Cardoso administration ends with another balance of payments crisis, which was empowered by the coming presidential elections. Yet, the solvency situation of the Brazilian economy have been improving since the 1999 successful floatation of the real, so that I believe that, adopted a policy that deepens fiscal adjustment, while lowers the interest rate, and avoids new evaluation of the real, the country will eventually be able to avoid default.

Instantes líricos de revelação: a narrativa poética em A paixão segundo G. H

Coordenação de Aperfeiçoamento de Pessoal de Nível Superior (CAPES)

Balancing Exploration and Exploitation: A New Algorithm for Active Machine Learning

Active machine learning algorithms are used when large numbers of unlabeled examples are available and getting labels for them is costly (e.g. requiring consulting a human expert). Many conventional active learning algorithms focus on refining the decision boundary, at the expense of exploring new regions that the current hypothesis misclassifies. We propose a new active learning algorithm that balances such exploration with refining of the decision boundary by dynamically adjusting the probability to explore at each step. Our experimental results demonstrate improved performance on data sets that require extensive exploration while remaining competitive on data sets that do not. Our algorithm also shows significant tolerance of noise.

Challenging market conventions:supermarket diversification and consumer resistance in children's apparel purchases

Purpose – In the UK, while fashion apparel purchasing is available to the majority of consumers, the main supermarkets seem – rather against the odds and market conventions – to have created a new, socially-acceptable and legitimate, apparel market offer for young children. This study aims to explore parental purchasing decisions on apparel for young children (below ten years old) focusing on supermarket diversification into apparel and consumer resistance against other traditional brands. Design/methodology/approach – Data collection adopted a qualitative research mode: using semi-structured interviews in two locations (Cornwall Please correct and check againand Glasgow), each with a Tesco and ASDA located outside towns. A total of 59 parents participated in the study. Interviews took place in the stores, with parents seen buying children fashion apparel. Findings – The findings suggest that decisions are based not only on functionality (e.g. convenience, value for money, refund policy), but also on intuitive factors (e.g. style, image, quality) as well as broader processes of consumption from parental boundary setting (e.g. curbing premature adultness). Positive consumer resistance is leading to a re-drawing of the cultural boundaries of fashion. In some cases, concerns are expressed regarding items that seem too adult-like or otherwise not as children's apparel should be. Practical implications – The paper highlights the increasing importance of browsing as a modern choice practice (e.g. planned impulse buying, sanctuary of social activity). Particular attention is given to explaining why consumers positively resist buying from traditional label providers and voluntarily choose supermarket clothing ranges without any concerns over their children wearing such garments. Originality/value – The paper shows that supermarket shopping for children's apparel is now firmly part of UK consumption habits and choice. The findings provide theoretical insights into the significance of challenging market conventions, parental cultural boundary setting and positive resistance behaviour.

Strategists' reactions and resistance towards forces of inclusion:soothing the anxiety of marketing (non-) influence

The firm is becoming more and more inclusive in its conception. And yet, marketing studies point to the same overwhelming conclusion that marketing, marketing departments and marketers are being increasingly 'pushed out' - excluded. We argue that where and when inclusion-exclusion intersect in the practice of strategic marketing is important, not least because their powerful boundary-setting and spanning roles have a determinant effect on the places and spaces, within which marketing strategists are (counter-) mobilized. This paper provides new insights relating to the contradictory forces existing around inclusion-exclusion in corporate strategizing. A further aim is to present the position of marketing (non-) influence within this context. The paper provides a unique theoretical contribution by illustrating some of the contradictions, struggles and activities that make the theoretical shift towards strategic inclusivity unstable, partial and by no means inevitable. A further contribution is a linking of this broader strategic debate, with anxieties over the influence of marketing in corporate strategizing. This leads to a discussion of the various ways that marketing research can sooth the anxiety of influence on multiple fronts via: understanding agency and strategic action; shaping marketing curriculum development; and, reconsidering the spatial dimensions of marketing influence. © 2010 Taylor & Francis.

Carrier Selection for Multi-commodity Flow Optimization in Cooperative Environments

Part 18: Optimization in Collaborative Networks

Asymmetries in Adversarial Settings

Even without formal guarantees of their effectiveness, adversarial attacks against Machine Learning models frequently fool new defenses. We identify six key asymmetries that contribute to this phenomenon and formulate four guidelines to build future-proof defenses by preventing such asymmetries. We also prove that attacking a classifier is NP-complete, while defending from such attacks is Sigma_2^P-complete. We then introduce Counter-Attack (CA), an asymmetry-free metadefense that determines whether a model is robust on a given input by estimating its distance from the decision boundary. Under specific assumptions CA can provide theoretical detection guarantees. Additionally, we prove that while CA is NP-complete, fooling CA is Sigma_2^P-complete. Even when using heuristic relaxations, we show that our method can reliably identify non-robust points. As part of our experimental evaluation, we introduce UG100, a new dataset obtained by applying a provably optimal attack to six limited-scale networks (three for MNIST and three for CIFAR10), each trained in three different manners.

Clinical and laboratory predictors of imported malaria in an outpatient setting: an aid to medical decision making in returning travellers with fever

Si le tableau clinique évoque une malaria et que le résultat des examens parasitologiques n?est pas disponible ou est négatif, le praticien n?a pas d?information basée sur l?évidence pour savoir s?il doit donner ou non un traitement présomptif. Afin d?identifier les facteurs cliniques et paracliniques prédictifs d?une parasitémie à Plasmodium, nous avons mené une étude prospective chez les voyageurs ou migrants en provenance d?une zone tropicale ou subtropicale et qui consultaient pour de la fièvre. Le questionnaire comprenait 49 items explorant les données démographiques, les caractéristiques du voyage, les éléments de l?anamnèse et de l?examen clinique ainsi que les résultats de laboratoire. 336 sujets avec données complètes ont été recrutés (97 patients atteints de malaria et 239 contrôles avec fièvre et examen parasitologique négatif). L?analyse de régression multivariée a permis d?identifier les facteurs prédictifs de maiaria suivants : prophylaxie inadéquate, sudations, absence de douleur abdominale, température )38"C, mauvais état général, splénomégalie, compte leucocytaire (1 O x 1 03/L, plaquettes ~ 1 5 0 x l 03/L, taux d?hémoglobine <12 g/dL et éosinophiles (5%. La présence d?une splénomégalie avait le coefficient de probabilité positif pour un diagnostic de malaria le plus élevé (1 3.6) ; venait ensuite la présence d?une thrombopénie (1 1 .O). Dans le contexte de la consultation ambulatoire de la Policlinique Médicale Universitaire (prévalence de malaria de 29%), la probabilité post- test d?avoir un examen parasitologique positif était de 85% pour la splénomégalie et de 82% pour la thrombopénie. Même si le seuil thérapeutique n?est pas absolument défini, il semble raisonnable d?envisager un traitement présomptif lorsque la probabilité post- test est >80%. Si le médecin est réticent à administrer un traitement sans documentation parasitologique, il devrait au moins se retenir d?entreprendre d?autres investigations coûteuses, et plutôt répéter l?examen parasitologique après 12-24 heures.