153 research outputs found
Continuous-Time Stochastic Games with Time-Bounded Reachability.
No full textWe study continuous-time stochastic games with time-bounded reachability objectives. We show that each vertex in such a game has a value (i.e., an equilibrium probability), and we classify the conditions under which optimal strategies exist. Finally we show how to compute optimal strategies in finite uniform games, and how to compute ε-optimal strategies in finitely-branching games with bounded rates (for finite games, we provide detailed complexity estimations). © Brazdil, Forejt, Krčál, Kretínsky, Kučera
Directional functional coupling of cerebral rhythms between anterior cingulate and dorsolateral prefrontal areas during rare stimuli: a directed transfer function analysis of human depth EEG signal.
No full textWhat is the neural substrate of our capability to properly react to changes in the environment? It can be hypothesized that the anterior cingulate cortex (ACC) manages repetitive stimuli in routine conditions and alerts the dorsolateral prefrontal cortex (PFC) when stimulation unexpectedly changes. To provide evidence in favor of this hypothesis, intracerebral stereoelectroencephalographic (SEEG) data were recorded from the anterior cingulate and dorsolateral PFC of eight epileptic patients in a standard Visual oddball task during presurgical monitoring. Two types of stimuli (200 ms duration) such as the letters O (frequent stimuli; 80% of probability) and X (rare stimuli) were presented in random order, with an interstimulus interval between 2 and 5 s. Subjects had to mentally count the rare (target) stimuli and to press a button with their dominant hand as quickly and accurately as possible. EEG frequency bands of interest here 0 (4-8 Hz), alpha(8-12 Hz), beta(14-30 Hz), and gamma(30-45 Hz). The directionality of the information flux within the EEG rhythms was indexed by a directed transfer function (DTF). The results showed that compared with the frequent stimuli, the target stimuli induced a statistically significant increase of DTF values from the anterior cingulate to the dorsolateral PFC at the 0 rhythms (P < 0.01). These results provide support to the hypothesis that ACC directly or indirectly affects the oscillatory activity of dorsolateral PFC by a selective frequency code under typical oddball conditions. Hum Brain Mapp 30:138-146,2009. (c) 2007 Wiley-Liss, Inc. RI Brazdil, Milan/D-6836-2012; Roman, Robert/E-3337-2012; Bares, Martin/E-3700-201
Five centuries of Central European temperature extremes reconstructed from tree-ring density and documentary evidence
No full textFuture climate change will likely influence the frequency and intensity of weather extremes. As such events are by definition rare, long records are required to understand their characteristics, drivers, and consequences on ecology and society. Herein we provide a unique perspective on regional-scale temperature extremes over the past millennium, using three tree-ring maximum latewood density (MXD) chronologies from higher elevations in the European Alps. We verify the tree-ring-based extremes using documentary evidences from Switzerland, the Czech Republic, and Central Europe that allowed the identification of 44 summer extremes over the 1550-2003 period. These events include cold temperatures in 1579, 1628, 1675, and 1816, as well as warm ones in 1811 and 2003. Prior to 1550, we provide new evidence for cold (e.g., 1068 and 1258) and warm (e.g., 1333) summers derived from the combined MXD records and thus help to characterize high-frequency temperature variability during medieval times. Spatial coherence of the reconstructed extremes is found over Switzerland, with most signatures even extending across Central Europe. We discuss potential limitations of the tree-ring and documentary archives, including the (i) ability of MXD to particularly capture extremely warm temperatures, (ii) methodological identification and relative definition of extremes, and (iii) placement of those events in the millennium-long context of low-frequency climate change. © 2010 Elsevier B.V
Evaluating the correlation between objective rule interestingness measures and real human interest
No full textIn the last few years, the data mining community has proposed a number of objective rule interestingness measures to select the most interesting rules, out of a large set of discovered rules. However, it should be recalled that objective measures are just an estimate of the true degree of interestingness of a rule to the user, the so-called real human interest. The latter is inherently subjective. Hence, it is not clear how effective, in practice, objective measures are. More precisely, the central question investigated in this paper is: “how effective objective rule interestingness measures are, in the sense of being a good estimate of the true, subjective degree of interestingness of a rule to the user?” This question is investigated by extensive experiments with 11 objective rule interestingness measures across eight real-world data sets
Comparing Strategies of Collaborative Networks for R&D: an agent-based study
No full textIn this work we analyze the evolving dynamics of different collaboration network strategies that emerge from the creation and diffusion of knowledge. In addition, we aim at describing their most relevant network properties over time. An evolutionary economic approach has been adopted by avoiding the profit-maximization behavior of firms and introducing decision rules that are applied routinely. A Multi-Agent Model with cognitive attributes where agents learn to make their own decisions has been developed. Firms (the agents) can collaborate and create networks for Research and Development (R&D) purposes. We have compared five collaboration strategies (A - Peer-to-Peer complementariness, B –Concentration process, C –Reinforcement Strategy, D - Virtual Collaboration Networks and E - Virtual Cooperation Networks) that were defined on the basis of literature and on empirical evidence. Strategies are introduced exogenously in the simulation. The aims of this paper are threefold: (i) to analyze the importance of the networking effects; (ii) to test the differences among collaboration strategies; and, finally, (iii) to verify the effect of learning. It has been possible to conclude that profit is associated with higher stock of knowledge and with smaller network diameter. In addition, concentration strategies are more profitable and more efficient in transmitting knowledge through the network. These processes reinforce the stock of knowledge and the profit of the firms located in the centers of the networks. Such dynamics is supported by the learning mechanism that generates a kind of collective cognition: in fact, if more firms connect to a particular network, then the center of the network is reinforced, producing feedbacks to all nodes.Collaborative networks, Multi-Agent System, Collaboration Strategies, Stock of Knowledge
Framework para descoberta científica suportada por interação híbrida homem-máquina
No full textDissertação submetida à UNIVERSIDADE DE TRÁS-OS-MONTES E ALTO DOURO para obtenção do grau de MESTRE em Engenharia InformáticaCompreender as interações nas comunidades científicas e as suas colaborações, tornou-se indispensável para a investigação propriamente dita. Assim sendo, a medição da similaridade entre documentos científicos poderá auxiliar os investigadores na identificação de grupos com interesses semelhantes, promovendo a colaboração e o reforço das ligações entre a academia e a indústria. Com este propósito, procura-se avaliar o desempenho de abordagens híbridas na medição de similaridade entre pares de documentos, atrav´es da complementaridade de resultados alcançados por crowd participants e algoritmos de inteligência artificial. Esta dissertação apresenta, assim, uma framework que integra dois componentes sequenciais, contendo dois tipos de processos destinados a perceber de que forma os sistemas que envolvem algoritmos computacionais podem colaborar com seres humanos, na medição de similaridade entre documentos científicos. A primeira componente envolve um conjunto de abordagens de Natural Language Processing (Processamento de Linguagem Natural) ou NLP e de Text Mining, na qual ´e utilizada a medida TFIDF e o modelo de representação Bidirectional Encoder Representation from Transformers (BERT). A segunda componente, consiste numa campanha de crowdsourcing, na qual os participantes (crowd participants) terão de indicar se os documentos cient´ıficos em causa são, ou não, da mesma autoria. A utilização de processos de crowdsourcing nas situações em que algoritmos automáticos não fornecem resultados satisfatórios, faculta uma visão preliminar na deteção de contribuições importantes da cooperação Human-AI. Partindo deste pressuposto, preconiza-se a medição de similaridade entre documentos científicos, visando alcançar um melhor suporte à decisão baseado num worflow híbrido. Assim sendo, acredita-se que os investigadores podem ser melhor informados sobre potenciais colaboradores, recorrendo a mecanismos híbridos de Human-AI baseados no conteúdo das suas publicações científicas.Understanding the intellectual landscape of scientific communities and their collaborations has become an indispensable part of research per se. In this regard, measuring similarities among scientific documents can help researchers to identify groups with similar interests as a basis for strengthening collaboration and university-industry linkages. To this end, we intend to evaluate the performance of hybrid crowd-computing methods in measuring the similarity between document pairs by comparing the results achieved by crowds and artificial intelligence (AI) algorithms. That said, this dissertation presents a framework constituted by two sequential components that contain two types of experiments to illustrate some issues in calculating how similar an automatic solution is to a given ground truth. The first component involves a set of natural language processing (NLP) processes in which we used the TF-IDF measure and the Bidirectional Encoder Representation from Transformers (BERT) model. For the next component, we created a crowdsourcing campaign consisting of four human intelligence tasks (HITs) in which the participants had to indicate whether or not a set of papers belonged to the same author. The use of crowdsourcing processes in situations where automatic algorithms do not provide satisfactory results provides preliminary insights into detecting major contributions from human-AI cooperation at similarity calculation in order to achieve better decision support and the advantage of a hybrid workflow system in this matter. We believe that in this case decision makers can be better informed about potential collaborators based on content-based insights enhanced by hybrid humanAI mechanisms
Generating dynamic higher-order Markov models in web usage mining
No full textMarkov models have been widely used for modelling users’ web navigation behaviour. In previous work we have presented a dynamic clustering-based Markov model that accurately represents second-order transition probabilities given by a collection of navigation sessions. Herein, we propose a generalisation of the method that takes into account higher-order conditional probabilities. The method makes use of the state cloning concept together with a clustering technique to separate the navigation paths that reveal differences in the conditional probabilities. We report on experiments conducted with three real world data sets. The results show that some pages require a long history to understand the users choice of link, while others require only a short history. We also show that the number of additional states induced by the method can be controlled through a probability threshold parameter
Ensembles of balanced nested dichotomies for multi-class problems
No full textA system of nested dichotomies is a hierarchical decomposition of a multi-class problem with c classes into c−1 two-class problems and can be represented as a tree structure. Ensembles of randomly generated nested dichotomies have proven to be an effective approach to multi-class learning problems [1]. However, sampling trees by giving each tree equal probability means that the depth of a tree is limited only by the number of classes, and very unbalanced trees can negatively affect runtime.
In this paper, we investigate two approaches to building balanced nested dichotomies—class-balanced nested dichotomies and data-balanced nested dichotomies—and evaluate them in the same ensemble setting. Using C4.5 decision trees as the base models, we show that both approaches can reduce runtime with little or no effect on accuracy, especially on problems with many classes. We also investigate the effect of caching models when building ensembles of nested dichotomies
Statistical Relational Learning: An Inductive Logic Programming Perspective
No full textstatus: Publishe
- …
