• European Marine Science
• Open Access close
• Conference object close
• arXiv.org e-Print Archive close

Animals form groups for many reasons but there are costs andbenefit associated with group formation. One of the benefits is collectivememory. In groups on the move, social interactions play a crucialrole in the cohesion and the ability to make consensus decisions.When migrating from spawning to feeding areas fish schools need toretain a collective memory of the destination site over thousand ofkilometers and changes in group formation or individual preferencecan produce sudden changes in migration pathways. We propose amodelling framework, based on stochastic adaptive networks, thatcan reproduce this collective behaviour. We assume that three factorscontrol group formation and school migration behaviour: the intensityof social interaction, the relative number of informed individualsand the preference that each individual has for the particular migrationarea. We treat these factors independently and relate the individuals’preferences to the experience and memory for certain migrationsites. We demonstrate that removal of knowledgable individualsor alteration of individual preference can produce rapid changes ingroup formation and collective behavior. For example, intensive fishingtargeting the migratory species and also their preferred prey canreduce both terms to a point at which migration to the destinationsites is suddenly stopped. The conceptual approaches represented byour modelling framework may therefore be able to explain large-scalechanges in fish migration and spatial distribution

Spectral characterization of Super-Earth atmospheres for planets orbiting in the Habitable Zone of M-dwarf stars is a key focus in exoplanet science. A central challenge is to understand and predict the expected spectral signals of atmospheric biosignatures (species associated with life). Our work applies a global-mean radiative-convective-photochemical column model assuming a planet with an Earth-like biomass and planetary development. We investigated planets with gravities of 1g and 3g and a surface pressure of one bar around central stars with spectral classes from M0 to M7. The spectral signals of the calculated planetary scenarios have been presented by Rauer et al. (2011). The main motivation of the present work is to perform a deeper analysis of the chemical processes in the planetary atmospheres. We apply a diagnostic tool, the Pathway Analysis Program, to shed light on the photochemical pathways that form and destroy biosignature species. Ozone is a potential biosignature for complex- life. An important result of our analysis is a shift in the ozone photochemistry from mainly Chapman production (which dominates in the terrestrial stratosphere) to smog-dominated ozone production for planets in the Habitable Zone of cooler (M5-M7)-class dwarf stars. This result is associated with a lower energy flux in the UVB wavelength range from the central star, hence slower planetary atmospheric photolysis of molecular oxygen, which slows the Chapman ozone production. 53 pages, 7 figures, manuscript accepted at "Astrobiology" Journal

Non-blind deconvolution aims to restore a sharp image from its blurred counterpart given an obtained kernel. Existing deep neural architectures are often built based on large datasets of sharp ground truth images and trained with supervision. Sharp, high quality ground truth images, however, are not always available, especially for biomedical applications. This severely hampers the applicability of current approaches in practice. In this paper, we propose a novel non-blind deconvolution method that leverages the power of deep learning and classic iterative deconvolution algorithms. Our approach combines a pre-trained network to extract deep features from the input image with iterative Richardson-Lucy deconvolution steps. Subsequently, a zero-shot optimisation process is employed to integrate the deconvolved features, resulting in a high-quality reconstructed image. By performing the preliminary reconstruction with the classic iterative deconvolution method, we can effectively utilise a smaller network to produce the final image, thus accelerating the reconstruction whilst reducing the demand for valuable computational resources. Our method demonstrates significant improvements in various real-world applications non-blind deconvolution tasks.

The DESERT Underwater framework (http://desert-underwater.dei.unipd.it/), originally designed for simulating and testing underwater acoustic networks in sea trials, has recently been extended to support real payload data transmission through underwater multimodal networks. Specifically, the new version of the framework is now able to transmit data in real time through the EvoLogics S2C low-rate and high-rate acoustic modems, the SmartPORT low-cost acoustic underwater modem prototype (AHOI) for IoT applications, as well as Ethernet, surface WiFi, and the BlueComm optical modem. The system can also be tested in the lab by employing a simulated channel, and the EvoLogics S2C DMAC Emulator (DMACE) Comment: 5 pages, 6 figures

Detecting novel objects without class information is not trivial, as it is difficult to generalize from a small training set. This is an interesting problem for underwater robotics, as modeling marine objects is inherently more difficult in sonar images, and training data might not be available apriori. Detection proposals algorithms can be used for this purpose but usually requires a large amount of output bounding boxes. In this paper we propose the use of a fully convolutional neural network that regresses an objectness value directly from a Forward-Looking sonar image. By ranking objectness, we can produce high recall (96 %) with only 100 proposals per image. In comparison, EdgeBoxes requires 5000 proposals to achieve a slightly better recall of 97 %, while Selective Search requires 2000 proposals to achieve 95 % recall. We also show that our method outperforms a template matching baseline by a considerable margin, and is able to generalize to completely new objects. We expect that this kind of technique can be used in the field to find lost objects under the sea. Comment: European Conference on Mobile Robots 2019

There is an emerging trend of embedding knowledge graphs (KGs) in continuous vector spaces in order to use those for machine learning tasks. Recently, many knowledge graph embedding (KGE) models have been proposed that learn low dimensional representations while trying to maintain the structural properties of the KGs such as the similarity of nodes depending on their edges to other nodes. KGEs can be used to address tasks within KGs such as the prediction of novel links and the disambiguation of entities. They can also be used for downstream tasks like question answering and fact-checking. Overall, these tasks are relevant for the semantic web community. Despite their popularity, the reproducibility of KGE experiments and the transferability of proposed KGE models to research fields outside the machine learning community can be a major challenge. Therefore, we present the KEEN Universe, an ecosystem for knowledge graph embeddings that we have developed with a strong focus on reproducibility and transferability. The KEEN Universe currently consists of the Python packages PyKEEN (Python KnowlEdge EmbeddiNgs), BioKEEN (Biological KnowlEdge EmbeddiNgs), and the KEEN Model Zoo for sharing trained KGE models with the community.

The next generation multi-cubic-kilometre water Cherenkov neutrino telescope will be build in the Mediterranean Sea. This telescope, called KM3NeT, is currently entering a first construction phase. The KM3NeT research infrastructure will comprise 690 so-called Detection Units in its final design which will be anchored to the sea bed and held upright by submerged floats. The positions of these Detection Units, several hundred metres in length, and their attached Optical Modules for the detection of Cherenkov light have to be monitored continously to provide the telescope with its desired pointing precision. A standard way to do this is the utilisation of an acoustic positioning system using emitters at fixed positions and receivers distributed along the Detection Units. The KM3NeT neutrino telescope comprises a custom-made acoustic positioning system with newly designed emitters attached to the anchors of the Detection Units and custom-designed receivers attached to the Detection Units. This article describes an approach for a receiver and its performance. The proposed Opto-Acoustical Modules combine the optical sensors for the telescope with the acoustical sensors necessary for the positioning of the module itself. This combination leads to a compact design suited for an easy deployment of the numerous Detection Units. Furthermore, the instrumented volume can be used for scientific analyses such as marine science and acoustic particle detection. 5 pages, 5 figures. To be published in: AIP Conference proceedings of the "VLVnT13 - Very Large Volume Neutrino Telescope Workshop (2013)"