Radboud University Nijmegen

About: Radboud University Nijmegen is a education organization based out in Nijmegen, Gelderland, Netherlands. It is known for research contribution in the topics: Population & Randomized controlled trial. The organization has 35417 authors who have published 83035 publications receiving 3285064 citations. The organization is also known as: Catholic University of Nijmegen & Radboud University.

Task clustering and gating for bayesian multitask learning

Abstract: Modeling a collection of similar regression or classification tasks can be improved by making the tasks 'learn from each other'. In machine learning, this subject is approached through 'multitask learning', where parallel tasks are modeled as multiple outputs of the same network. In multilevel analysis this is generally implemented through the mixed-effects linear model where a distinction is made between 'fixed effects', which are the same for all tasks, and 'random effects', which may vary between tasks. In the present article we will adopt a Bayesian approach in which some of the model parameters are shared (the same for all tasks) and others more loosely connected through a joint prior distribution that can be learned from the data. We seek in this way to combine the best parts of both the statistical multilevel approach and the neural network machinery. The standard assumption expressed in both approaches is that each task can learn equally well from any other task. In this article we extend the model by allowing more differentiation in the similarities between tasks. One such extension is to make the prior mean depend on higher-level task characteristics. More unsupervised clustering of tasks is obtained if we go from a single Gaussian prior to a mixture of Gaussians. This can be further generalized to a mixture of experts architecture with the gates depending on task characteristics. All three extensions are demonstrated through application both on an artificial data set and on two real-world problems, one a school problem and the other involving single-copy newspaper sales.

Importance of Mangroves, Seagrass Beds and the Shallow Coral Reef as a Nursery for Important Coral Reef Fishes, Using a Visual Census Technique

Abstract: The nursery function of various biotopes for coral reef fishes was investigated on Bonaire, Netherlands Antilles. Length and abundance of 16 commercially important reef fish species were determined by means of visual censuses during the day in six different biotopes: mangrove prop-roots ( Rhizophora mangle ) and seagrass beds ( Thalassia testudinum ) in Lac Bay, and four depth zones on the coral reef (0 to 3 m, 3 to 5 m, 10 to 15 m and 15 to 20 m). The mangroves, seagrass beds and shallow coral reef (0 to 3 m) appeared to be the main nursery biotopes for the juveniles of the selected species. Mutual comparison between biotopes showed that the seagrass beds were the most important nursery biotope for juvenile Haemulon flavolineatum , H. sciurus , Ocyurus chrysurus , Acanthurus chirurgus and Sparisoma viride , the mangroves for juvenile Lutjanus apodus , L. griseus , Sphyraena barracuda and Chaetodon capistratus , and the shallow coral reef for juvenile H. chrysargyreum , L. mahogoni , A. bahianus and Abudefduf saxatilis . Juvenile Acanthurus coeruleus utilized all six biotopes, while juvenile H. carbonarium and Anisotremus surinamensis were not observed in any of the six biotopes. Although fishes showed a clear preference for a specific nursery biotope, most fish species utilized multiple nursery biotopes simultaneously. The almost complete absence of juveniles on the deeper reef zones indicates the high dependence of juveniles on the shallow water biotopes as a nursery. For most fish species an (partial) ontogenetic shift was observed at a particular life stage from their (shallow) nursery biotopes to the (deeper) coral reef. Cluster analyses showed that closely related species within the families Haemulidae, Lutjanidae and Acanthuridae, and the different size classes within species in most cases had a spatial separation in biotope utilization.

Magnetic Correlations at Graphene Edges : Basis for Novel Spintronics Devices

Abstract: Magnetic zigzag edges of graphene are considered as a basis for novel spintronics devices despite the fact that no true long-range magnetic order is possible in one dimension. We study the transverse and longitudinal fluctuations of magnetic moments at zigzag edges of graphene from first principles. We find a high value for the spin wave stiffness D=2100 meV A2 and a spin-collinear domain wall creation energy E(dw)=114 meV accompanied by low magnetic anisotropy. Above the crossover temperature T(x) approximately 10 K, the spin correlation length xi proportional, variantT(-1) limits the long-range magnetic order to approximately 1 nm at 300 K while below T(x), it grows exponentially with decreasing temperature. We discuss possible ways of increasing the range of magnetic order and effects of edge roughness on it.

Climate change and unequal phenological changes across four trophic levels: constraints or adaptations?

Abstract: 1. Climate change has been shown to affect the phenology of many organisms, but interestingly these shifts are often unequal across trophic levels, causing a mismatch between the phenology of organisms and their food. 2. We consider two alternative hypotheses: consumers are constrained to adjust sufficiently to the lower trophic level, or prey species react more strongly than their predators to reduce predation. We discuss both hypotheses with our analyses of changes in phenology across four trophic levels: tree budburst, peak biomass of herbivorous caterpillars, breeding phenology of four insectivorous bird species and an avian predator. 3. In our long-term study, we show that between 1988 and 2005, budburst advanced (not significantly) with 0.17 d yr(-1), while between 1985 and 2005 both caterpillars (0.75 d year(-1)) and the hatching date of the passerine species (range for four species: 0.36-0.50 d year(-1)) have advanced, whereas raptor hatching dates showed no trend. 4. The caterpillar peak date was closely correlated with budburst date, as were the passerine hatching dates with the peak caterpillar biomass date. In all these cases, however, the slopes were significantly less than unity, showing that the response of the consumers is weaker than that of their food. This was also true for the avian predator, for which hatching dates were not correlated with the peak availability of fledgling passerines. As a result, the match between food demand and availability deteriorated over time for both the passerines and the avian predators. 5. These results could equally well be explained by consumers' insufficient responses as a consequence of constraints in adapting to climate change, or by them trying to escape predation from a higher trophic level, or both. Selection on phenology could thus be both from matches of phenology with higher and lower levels, and quantifying these can shed new light on why some organisms do adjust their phenology to climate change, while others do not.