• Other research productsclose
• European Commission close
• GB close
• GR close
• IL close
• English close

We consider the problem of decision making in the context of unknown Markov decision processes with finite state and action spaces. In a Bayesian reinforcement learning framework, we propose an optimistic posterior sampling strategy based on the maximization of state-action value functions of MDPs sampled from the posterior. First experiments are promising.

How far should countries go to encourage foreign investment? Jelena Dzankic writes that in a time of economic crisis, some countries in Europe are now seeking investment in exchange for citizenship. Assessing recent developments in Bulgaria, Hungary, Portugal and Ireland, she argues that despite the obvious financial benefits to such policies, they are not without risks. They may raise the potential for tax evasion and security issues, and could also reduce the relationship between the individual and the state to that of a business contract.

This policy report is based in Epinet WP2 and complements WP1 reporting on the EPINET Integrated Assessment Framework as a Tool for RRI . We present key findings from the empirical research we conducted and was designed to be an instrument of observation and reflexivity in reference to the interdisciplinary innovation assessment cases conducted as part of the Epinet project. In particular, we report on the procedural conditions in carrying out these cases as the basis on which our policy recommendations rest.

The development of phosphate sensors suitable for long-term in situ deployments in natural waters, is essential to improve our understanding of the distribution, fluxes, and biogeochemical role of this key nutrient in a changing ocean. Here, we describe the optimization of the molybdenum blue method for in situ work using a lab-on-chip (LOC) analyzer and evaluate its performance in the laboratory and at two contrasting field sites. The in situ performance of the LOC sensor is evaluated using hourly time-series data from a 56-day trial in Southampton Water (UK), as well as a month-long deployment in the subtropical oligotrophic waters of Kaneohe Bay (Hawaii, USA). In Kaneohe Bay, where phosphate concentrations were characteristic of the dry season (0.13 ± 0.03 μM, n = 704), the in situ sensor accuracy was 16 ± 12% and a potential diurnal cycle in phosphate concentrations was observed. In Southampton Water, the sensor data (1.02 ± 0.40 μM, n = 1,267) were accurate to ±0.10 μM relative to discrete reference samples. Hourly in situ monitoring revealed striking tidal and storm derived fluctuations in phosphate concentrations in Southampton Water that would not have been captured via discrete sampling. We show the impact of storms on phosphate concentrations in Southampton Water is modulated by the spring-neap tidal cycle and that the 10-fold decline in phosphate concentrations observed during the later stages of the deployment was consistent with the timing of a spring phytoplankton bloom in the English Channel. Under controlled laboratory conditions in a 250 L tank, the sensor demonstrated an accuracy and precision better than 10% irrespective of the salinity (0–30), turbidity (0–100 NTU), colored dissolved organic matter (CDOM) concentration (0–10mg/L), and temperature (5–20◦C) of the water (0.3–13 μM phosphate) being analyzed. This work demonstrates that the LOC technology is mature enough to quantify the influence of stochastic events on nutrient budgets and to elucidate the role of phosphate in regulating phytoplankton productivity and community composition in estuarine and coastal regimes. Refereed 14.A Nutrients TRL 8 Actual system completed and "mission qualified" through test and demonstration in an operational environment (ground or space) Manual (incl. handbook, guide, cookbook etc) Standard Operating Procedure

The Second Annual Meeting of the European Virus Bioinformatics Center (EVBC), held in Utrecht, Netherlands, focused on computational approaches in virology, with topics including (but not limited to) virus discovery, diagnostics, (meta-)genomics, modeling, epidemiology, molecular structure, evolution, and viral ecology. The goals of the Second Annual Meeting were threefold: (i) to bring together virologists and bioinformaticians from across the academic, industrial, professional, and training sectors to share best practice; (ii) to provide a meaningful and interactive scientific environment to promote discussion and collaboration between students, postdoctoral fellows, and both new and established investigators; (iii) to inspire and suggest new research directions and questions. Approximately 120 researchers from around the world attended the Second Annual Meeting of the EVBC this year, including 15 renowned international speakers. This report presents an overview of new developments and novel research findings that emerged during the meeting.

Spectroscopic studies of Galactic O and B stars show that many stars with masses above 8 M$_{\odot}$ are observed in the HR diagram just beyond the Main-Sequence (MS) band predicted by stellar models computed with a moderate overshooting. This may be an indication that the convective core sizes in stars in the upper part of the HR diagram are larger than predicted by these models. Combining stellar evolution models and spectroscopic parameters derived for a large sample of Galactic O and B stars, including brand new information about their projected rotational velocities, we reexamine the question of the convective core size in MS massive stars. We confirm that for stars more massive than about 8 M$_{\odot}$, the convective core size at the end of the MS phase increases more rapidly with the mass than in models computed with a constant step overshoot chosen to reproduce the main sequence width in the low mass range (around 2 M$_{\odot}$). This conclusion is valid for both the cases of non-rotating models and rotating models either with a moderate or a strong angular momentum transport. The increase of the convective core mass with the mass obtained from the TAMS position is, however, larger than the one deduced from the surface velocity drop for masses above about 15 M$_{\odot}$. Although observations available at the moment cannot decide what is the best choice between the core sizes given by the TAMS and the velocity drop, we discuss different methods to get out of this dilemma. At the moment, comparisons with eclipsing binaries seem to favor the solution given by the velocity drop. While we confirm the need for larger convective cores at higher masses, we find tensions in-between different methods for stars more massive than 15 M$_{\odot}$. The use of single-aged stellar populations (non-interacting binaries or stellar clusters) would be a great asset to resolve this tension.

Automatic summarization of reader comments in on-line news is an extremely challenging task and a capability for which there is a\ud clear need. Work to date has focussed on producing extractive summaries using well-known techniques imported from other areas of\ud language processing. But are extractive summaries of comments what users really want? Do they support users in performing the sorts\ud of tasks they are likely to want to perform with reader comments? In this paper we address these questions by doing three things. First,\ud we offer a specification of one possible summary type for reader comment, based on an analysis of reader comment in terms of issues\ud and viewpoints. Second, we define a task-based evaluation framework for reader comment summarization that allows summarization\ud systems to be assessed in terms of how well they support users in a time-limited task of identifying issues and characterising opinion on\ud issues in comments. Third, we describe a pilot evaluation in which we used the task-based evaluation framework to evaluate a prototype\ud reader comment clustering and summarization system, demonstrating the viability of the evaluation framework and illustrating the sorts\ud of insight such an evaluation affords.