Search

For the 24th Machine Learning for Signal Processing competition, participants were asked to automatically diagnose schizophrenia using multimodal features derived from MRI scans. The objective of the classification task was to achieve the best possible schizophrenia diagnosis prediction based only on the multimodal features derived from brain MRI scans. A total of 2087 entries from 291 participants with active Kaggle.com accounts were made. Each participant developed a classifier, with optional feature selection, that combined functional and structural magnetic resonance imaging features. Here we review details about the competition setup, the winning strategies, and provide basic analyses of the submitted entries. We conclude with a discussion of the advances made to the neuroimaging and machine learning fields. Index Terms— Competition, MRI, Schizophrenia, FNC, SBM 1. INTRODUCTION Schizophrenia is a severe and disabling mental illness which has no well-established, non-invasive diagnosis biomarker. Currently, due to its symptom overlap with other mental illnesses (like bipolar disorder) it is diagnosed subjectively only, by process of elimination. In order to broadly explore the current limits of automatic diagnosis of schizophrenia, this year’s machine learning competition was entitled The MLSP 2014 Schizophrenia Classification Challenge, which occurred from June 5 through July 20, 2014. For the second consecutive year, the competition was hosted on Kaggle.com. The response from neuroscience and machine learning communities was fantastic, including 291 participants in 245 teams, and an impressive 2087 valid entries. The competition was sponsored by Kaggle.com, who provided their hosting services, and Springer, who provided monetary rewards to the winning participants. In total, 611 entries from 151 participants outperformed the provided competition benchmark in all performance measures considered here. The three winning strategies used, in ranking order, Gaussian process (GP) [1], Gaussian kernel support vector machine (SVM) [2], and distance weighted discrimination (DWD) [3] classifiers, respectively.

Multimodal fusion is becoming more common as it proves to be a powerful approach to identify complementary information from multimodal datasets. However, simulation of joint information is not straightforward. Published approaches mostly employ limited, provisional designs that often break the link between the model assumptions and the data for the sake of demonstrating properties of fusion techniques. This work introduces a new approach to synthetic data generation which allows full-compliance between data and model while still representing realistic spatiotemporal features in accordance with the current neuroimaging literature. The focus is on the simulation of joint information for the verification of stochastic linear models, particularly those used in multimodal data fusion of brain imaging data. Our first goal is to obtain a benchmark ground-truth in which estimation errors due to model mismatch are minimal or none. Then we move on to assess how estimation is affected by gradually increasing model discrepancies toward a more realistic dataset. The key aspect of our approach is that it permits complete control over the type and level of model mismatch, allowing for more educated inferences about the limitations and caveats of select stochastic linear models. As a result, impartial comparison of models is possible based on their performance in multiple different scenarios. Our proposed method uses the commonly overlooked theory of copulas to enable full control of the type and level of dependence/association between modalities, with no occurrence of spurious multimodal associations. Moreover, our approach allows for arbitrary single-modality marginal distributions for any fixed choice of dependence/association between multimodal features. Using our simulation framework, we can rigorously challenge the assumptions of several existing multimodal fusion approaches. Our study brings a new perspective to the problem of simulating multimodal data that can be used for ground-truth verification of various stochastic multimodal models available in the literature, and reveals some important aspects that are not captured or are overlooked by ad hoc simulations that lack a firm statistical motivation.

The recent discovery of pulsating subdwarf B (sdB) stars at the South African Astronomical Observatory (Kilkenny et al. 1997; Koen et al. 1997; Stobie et al. 1997; O'Donoghue et al. 1997) has opened a brand new avenue in the field of asteroseismology. The study of these pulsators (dubbed EC14026 stars, after the prototype) offers the exciting possibility of exploiting the full power of asteroseismology to investigate the sdB phase of stellar evolution, one of the last frontiers in our general understanding of the history of stars.

Improving governance is central to improving results in human development. It is clear that money is not enough: improved outcomes from service delivery require better governance, including mechanisms for holding service providers accountable and appropriate incentives for performance. There is therefore a growing demand for indicators to measure how and whether these processes work, and how they affect health and education results. This paper makes the case for measuring governance policies and performance, and the quality of service delivery in health and education. It develops a framework for selecting and measuring a set of indicators and proposes options, drawing from new and innovative measurement tools and approaches. The paper proposes the adoption of a more systematic approach that will both facilitate the work of health and education policymakers and allow for cross-country comparisons and benchmarking.

According to World Bank policy, countries remain eligible to borrow from the International Bank for Reconstruction and Development until they are able to sustain long-term development without further recourse to Bank financing. Graduation from the Bank is not an automatic consequence of reaching a particular income level, but rather is supposed to be based on a determination of whether the country has reached a level of institutional development and capital-market access that enables it to sustain its own development process without recourse to Bank funding. This paper assesses how International Bank for Reconstruction and Development graduation policy operates in practice, investigating what income and non-income factors appear to have influenced graduation decisions in recent decades, based on panel data for 1982 through 2008. Explanatory variables include the per-capita income of the country, as well as measures of institutional development and market access that are cited as criteria by the graduation policy, and other plausible explanatory variables that capture the levels of economic development and vulnerability of the country. The authors find that the observed correlates of Bank graduation are generally consistent with the stated policy. Countries that are wealthier, more creditworthy, more institutionally developed, and less vulnerable to shocks are more likely to have graduated. Predicted probabilities generated by the model correspond closely to the actual graduation and de-graduation experiences of most countries (such as Korea and Trinidad and Tobago), and suggest that Hungary and Latvia may have graduated prematurely -- a prediction consistent with their subsequent return to borrowing from the Bank in the wake of the global financial crisis.

Quality management principles have been applied in a multidiscipline engineering design environment that emphasizes designing and building advanced technical facilities. The formulation of group objectives and annual goals and the alignment of improvement actions are discussed. The improvement process involves organizational, team, and individual actions supported by multi-tiered measurements and training. Both customer satisfaction and critical group performance measures have shown important improvements over the past 4 years.

The global financial crisis has not only dealt a major blow to the global economy, but also shaken confidence in economic management in the developed world and the economic models that guide it. The crisis has revealed major market failures, especially in the housing bubble and its transmission to the financial system, but also glaring state failures that propagated and exacerbated the crisis. Will the events of the past two years lead to major shifts in thinking about development economics, and should they? This paper assesses that question for several key domains of development thinking, including the market-state balance, macroeconomic management, globalization, development financing, and public spending. On the one hand, changed global circumstances and new awareness of vulnerability should lead to some policy changes, as developing countries take steps to reduce and buffer risks, including risks generated in developed countries. At the same time, the crisis should largely reinforce the Post-Washington Consensus on development that has emerged over the past decade -- a world view that aims to achieve private sector-driven growth but sees a facilitating role for the state, promotes engaging with the global economy in ways that advance development, and values pragmatism, experimentation, and evidence-based policymaking over ideology.

Expanding and improving basic education in developing countries requires, at a minimum, teachers who are present in the classroom and motivated to teach, but this essential input is often missing. This paper describes the findings of a series of recent World Bank and other studies on teacher absence and incentives for performance. Surprise school visits reveal that teachers are absent at high rates in countries such as India, Indonesia, Uganda, Ecuador, and Zambia, reducing the quality of schooling for children, especially in rural, remote, and poor areas. More broadly, poor teacher management and low levels of teacher accountability afflict many developing-country education systems. The paper presents evidence on these shortcomings, but also on the types of incentives, management, and support structures that can improve motivation and performance and reduce avoidable absenteeism. It concludes with policy options for developing countries to explore as they work to meet Education for All goals and improve quality.

Does private tutoring increase parental choice and improve student achievement, or does it exacerbate social inequalities and impose heavy costs on households, possibly without improving student outcomes? Private tutoring is now a major component of the education sector in many developing countries, yet education policy too seldom acknowledges or makes use of it. This survey of the literature examines the extent of private tutoring, identifies the factors that explain its growth, and analyzes its cost-effectiveness in improving student academic performance. It also presents a framework for assessing the efficiency and equity effects of tutoring. The results suggest that even taking equity concerns into account, tutoring can raise the effectiveness of the education system under certain reasonable assumptions. Guidance is offered for attacking corruption and other problems that diminish the benefits of private tutoring.

The acquisition of multiple imaging modalities on the same individual in brain imaging studies has become a common practice. In functional studies, often several different tasks are performed on the same person. This motivates an analysis method that directly looks for the joint (shared) information lying within multimodal datasets. The present work uses a data fusion framework called joint independent component analysis (jICA) to yield joint (multimodal), maximally independent components (ICs) which capture the joint information from multiple modalities and enable identification of brain imaging biomarkers. We thus propose the use of a divergence metric on the estimated group distributions as an optimization factor for this framework, thus characterizing the differences in the across-group distribution functions for each modality individually and jointly as well. Special attention is being devoted to the behavior aspects of the J- divergence and Alpha divergence (with alpha= 0.5) due to their metric property and optimality, respectively.

Private tutoring is now a major component of the education sector in many developing countries, yet education policy too seldom acknowledges and makes use of it. Various criticisms have been raised against private tutoring, most notably that it exacerbates social inequalities and may even fail to improve student outcomes. This paper surveys the literature for evidence on private tutoring-the extent of the tutoring phenomenon, the factors that explain its growth, and its cost-effectiveness in improving student academic performance. It also presents a framework for assessing the efficiency and equity effects of tutoring. It concludes that tutoring can raise the effectiveness of the education system under certain reasonable assumptions, even taking into account equity concerns, and it offers guidance for attacking corruption and other problems that diminish the contributions of the tutoring sector.

The acquisition of multiple brain imaging types for a given study is a very common practice. However these data are typically examined in separate analyses, rather than in a combined model. We propose a novel methodology to perform joint independent component analysis across image modalities, including structural MRI data, functional MRI activation data and EEG data, and to visualize the results via a joint histogram visualization technique. Evaluation of which combination of fused data is most useful is determined by using several information theoretic divergence measures. We demonstrate our method on a data set composed of functional MRI data from two tasks, structural MRI data, and EEG data collected on patients with schizophrenia and healthy controls. Our method provides a way to improve feature selection and even preprocessing. We show that combining data types can improve our ability to distinguish differences between groups.

We present an improved analytical expression for the x-ray dynamic structure factor from a dense plasma which includes the effects of weakly bound electrons. This result can be applied to describe scattering from low to moderate Z plasmas, and it covers the entire range of plasma conditions that can be found in inertial confinement fusion experiments, from ideal to degenerate up to moderately coupled systems. We use our theory to interpret x-ray scattering experiments from solid density carbon plasma and to extract accurate measurements of electron temperature, electron density and charge state. We use our experimental results to validate various equation-of-state models for carbon plasmas.

BACKGROUND: Current fibrinolytic therapies fail to achieve optimum reperfusion in many patients. Low-molecular-weight heparins and platelet glycoprotein IIb/IIIa inhibitors have shown the potential to improve pharmacological reperfusion therapy. We did a randomised, open-label trial to compare the efficacy and safety of tenecteplase plus enoxaparin or abciximab, with that of tenecteplase plus weight-adjusted unfractionated heparin in patients with acute myocardial infarction. METHODS: 6095 patients with acute myocardial infarction of less than 6 h were randomly assigned one of three regimens: full-dose tenecteplase and enoxaparin for a maximum of 7 days (enoxaparin group; n=2040), half-dose tenecteplase with weight-adjusted low-dose unfractionated heparin and a 12-h infusion of abciximab (abciximab group; n=2017), or full-dose tenecteplase with weight-adjusted unfractionated heparin for 48 h (unfractionated heparin group; n=2038). The primary endpoints were the composites of 30-day mortality, in-hospital reinfarction, or in-hospital refractory ischaemia (efficacy endpoint), and the above endpoint plus in-hospital intracranial haemorrhage or in-hospital major bleeding complications (efficacy plus safety endpoint). Analysis was by intention to treat. FINDINGS: There were significantly fewer efficacy endpoints in the enoxaparin and abciximab groups than in the unfractionated heparin group: 233/2037 (11.4%) versus 315/2038 (15.4%; relative risk 0.74 [95% CI 0.63-0.87], p=0.0002) for enoxaparin, and 223/2017 (11.1%) versus 315/2038 (15.4%; 0.72 [0.61-0.84], p