Search

International audience; Electroencephalography (EEG) is the most common non-invasive technique for measuring brain activity. It uses electrodes to capture the electric fields with high temporal resolution. However, analysing this data with machine learning presents several challenges, the most prominent one being the high inter/intra-subject variability (i.e. BCI inefficiency). One solution to tackle this challenge is to explore alternative features that capture additional types of information and help to better discriminate the subjects' mental state. For instance, complementary features that reflect interactions between brain areas, rather than relying solely on local measurements as proven efficient [1]. Functional Connectivity (FC) studies the interaction between pairs of electrodes in the spectral domain, embedding information in robust but distributed structures. To uncover those structures, graph-based approaches, such as Graph Neural Networks (GNN), demonstrated their ability to analyse complex networks, such as the interconnected nature of brain signals [2].This study proposes a GNN classifier method that combines FC information from various estimators like the Riemannian Covariance, Instantaneous and Imaginary Coherence with traditional feature space such as raw EEG features and spectral phase for each electrode for Brain-Computer Interfaces decoding. Preliminary results obtained with the High Gamma Dataset[3] show that the algorithm could successfully classify MI movements, but further experiments are necessary to evaluate the robustness of the intra-subject and cross-subject methods.

Job scheduling in high-performance computing platforms is a hard problem that involves uncertainties on both the job arrival process and their execution time. Users typically provide a loose upper bound estimate for job execution times that are hardly useful. Previous studies attempted to improve these estimates using regression techniques. Although these attempts provide reasonable predictions, they require a long period of training data. Furthermore, aiming for perfect prediction may be of limited use for scheduling purposes. In this work, we propose a simpler approach by classifying jobs as small or large and prioritizing the execution of small jobs over large ones. Indeed, small jobs are the most impacted by queuing delays but they typically represent a light load and incur a small burden on the other jobs. The classifier operates online and learns by using data collected over the previous weeks, facilitating its deployment and enabling fast adaptations to changes in workload characteristics. We evaluate our approach using four scheduling policies on six HPC platform workload traces. We show that: (i) incorporating such classification reduces the average bounded slowdown of jobs in all scenarios, and (ii) the obtained improvements are comparable, in most scenarios, to the ideal hypothetical situation where the scheduler would know the exact running time of jobs in advance.

No contexto de Cidades Inteligentes, utilizamos dados históricos da SPTrans para modelar os tempos de viagem de 20 linhas de ônibus da cidade de São Paulo como distribuições de probabilidade, analisando sua variabilidade. A distribuição que melhor caracterizou os tempos de viagem na maior parte dos casos foi a lognormal, indicando que há uma parcela de ônibus demorando muito mais do que a média. Analisamos também como cada um dos fatores constituintes do serviço (tamanho das linhas, número de paradas, tempo médio de viagem, velocidade média) se relacionavam entre si e afetavam a qualidade do serviço.

Tese (doutorado) - Universidade Federal de Santa Catarina, Centro Sócio-Econômico, Programa de Pós-Graduação em Contabilidade, Florianópolis, 2018. A Governança Corporativa (GC) oferece um conjunto de mecanismos de controle e de vinculação para mitigar o problema de agência. Este estudo tem como objetivo avaliar a relação entre os determinantes e os níveis de GC de empresas com ações negociadas na B3. Foi desenvolvido por meio de uma pesquisa documental com abordagem predominantemente quantitativa. A pesquisa tem como fundamento a Teoria da Agência e se baseia em um amplo levantamento de estudos que oferecem sustentação para o desenvolvimento da pesquisa, bem como para a avaliação das hipóteses estudadas. Foram utilizados dados de empresas integrantes do índice IBovespa referentes aos anos de 2015 e 2016. A amostra foi composta por 82 observações obtidas em relatórios financeiros de 41 empresas. Foram analisados os Formulários de Referência, Demonstrações Financeiras Padronizadas, Relatórios Anuais e de Sustentabilidade, Códigos de Conduta Ética, Políticas de Negociação de Valores Mobiliários, Estatutos Sociais, Comunicados ao Mercado referentes a substituições de auditores, listagem de ADRs disponibilizada pela JP Morgan, Websites de relações com investidores e o Setor de Atuação das empresas. Os dados foram avaliados com base em Estatísticas Descritivas, Teoria de Resposta ao Item (TRI) e Análise de Regressão Linear Múltipla. Por meio da TRI foi desenvolvida uma escala de avaliação das empresas de acordo com seus níveis de GC e a análise dos determinantes dos níveis de GC foi realizada mediante a aplicação da técnica de regressão. Os resultados permitem o estudo da estrutura de governança das empresas, bem como conhecer o grau de aplicação e efetividade dos fatores de GC apresentados pelas empresas. Os resultados oferecem também uma representação das dimensões que compõem a estrutura de governança das empresas, segundo a visão de nexus de contratos postulada pela Teoria da Agência. Por meio desta pesquisa é possível compreender quais mecanismos de controle e de vinculação são mais ou menos atendidos pelas empresas e como são atendidos. A escala de mensuração permite a comparação das diferentes empresas com base em um mesmo continuum. Nesta escala as empresas são classificadas nos níveis A, B, C e D, de acordo com a efetividade de aplicação dos mecanismos de controle e de vinculação. Ademais, o estudo permite avaliar os determinantes destes níveis de GC com base em características apresentadas pelas empresas. Os resultados da pesquisa sustentam que, dentre os atributos observáveis com base em informações contábeis, as empresas que possuem maiores níveis de GC são aquelas que com maiores oportunidades de investimentos e com menores níveis de alavancagem financeira. Dentre os atributos não contábeis, constatou-se que possuem maiores níveis de GC as empresas que atuam em setores regulamentados, que têm menor tempo de listagem e que emitem ADRs. Estes resultados contribuem para o desenvolvimento de conhecimentos acerca das estruturas de governança das empresas e oferecem também contribuição para o avanço dos demais estudos das áreas de Contabilidade, Finanças, Governança e áreas afins que buscam compreender os reflexos da GC. Abstract : Corporate Governance (CG) offers a set of control and alignment mechanisms to mitigate the agency problem. This study aims to evaluate the relationship between determinants and CG levels of companies with shares traded in B3. It was developed through archival research with a predominantly quantitative approach. The research is based on the Agency Theory and is based on a broad analysis of studies that provide support for the development of the research, as well as for the evaluation of the hypotheses studied. Companies data included in the IBovespa index for the years 2015 and 2016 were used. The sample consisted of 82 observations obtained from financial reports of 41 companies. The Reference Forms, Financial Statements, Annual and Sustainability Reports, Ethical Conduct Codes, Securities Trading Policies, Social Bylaws, Notice to the Market regarding auditors switch, JP Morgan s ADRs list, Investor Relations Websites and the Sector. The data were analyzed by Descriptive Statistics, Item Response Theory (IRT) and Multiple Linear Regression Analysis. Through TRI, a scale was used to evaluate the companies according to their CG levels and the determinants analysis of CG levels was performed using the regression. The results allow the study of the GC Structure, as well as the degree of application and effectiveness of the CG factors presented by companies. The results also provide a representation of the dimensions that make up the governance structure of companies, according to the nexus view of contracts postulated by the Agency Theory. This research allows to understand which control and alignment mechanisms are more or less attended to by the companies and how they are attended. The measurement scale allows the comparison of different companies based on the same continuum. In this scale the companies are classified in the levels A, B, C and D, according to the effectiveness of the application of control and alignment mechanisms. In addition, the study allows to evaluate the determinants of these CG levels based on characteristics presented by the companies. The results of the research support that, among the attributes observable based on accounting information, the companies that have higher levels of CG are those with greater investment opportunities and with lower levels of financial leverage. Among the non-accounting attributes, it was found that companies that operate in regulated sectors, which have less listing time and issue ADRs, have higher levels of CG. These results contribute to the development of knowledge about Corporate Governance Structures and also contribute to the advancement of other studies in the areas of Accounting, Finance, Governance and related areas that seek to understand the CG effects.

International audience; Dynamic scheduling of tasks in large-scale HPC platforms is normally accomplished using ad-hoc heuristics, based on task characteristics, combined with some backfilling strategy. Defining heuristics that work efficiently in different scenarios is a difficult task, specially when considering the large variety of task types and platform architectures. In this work, we present a methodology based on simulation and machine learning to obtain dynamic scheduling policies. Using simulations and a workload generation model, we can determine the characteristics of tasks that lead to a reduction in the mean slowdown of tasks in an execution queue. Modeling these characteristics using a nonlinear function and applying this function to select the next task to execute in a queue dramatically improved the mean task slowdown in synthetic workloads. When applied to real workload traces from highly different machines, these functions still resulted in important performance improvements, attesting the generalization capability of the obtained heuristics.

Humans’ and non-human animals’ ability to process time on the scale of milliseconds and seconds is essential for adaptive behaviour. A central question of how brains keep track of time is how specific temporal information across different sensory modalities is. In the present study, we show that encoding of temporal intervals in auditory and visual modalities are qualitatively similar. Human participants were instructed to reproduce intervals in the range from 750 ms to 1500 ms marked by auditory or visual stimuli. Our behavioural results suggest that, although participants were more accurate in reproducing intervals marked by auditory stimuli, there was a strong correlation in performance between modalities. Using multivariate pattern analysis in scalp EEG, we show that activity during late periods of the intervals was similar within and between modalities. Critically, we show that a multivariate pattern classifier was able to accurately predict the elapsed interval, even when trained on an interval marked by a stimulus of a different sensory modality. Taken together, our results suggest that, while there are differences in the processing of intervals marked by auditory and visual stimuli, they also share a common neural representation.