Search

A Federated Learning approach consists of creating an AI model from multiple data sources, without moving large amounts of data across to a central environment. Federated learning can be very useful in a tactical coalition environment, where data can be collected individually by each of the coalition partners, but network connectivity is inadequate to move the data to a central environment. However, such data collected is often dirty and imperfect. The data can be imbalanced, and in some cases, some classes can be completely missing from some coalition partners. Under these conditions, traditional approaches for federated learning can result in models that are highly inaccurate. In this paper, we propose approaches that can result in good machine learning models even in the environments where the data may be highly skewed, and study their performance under different environments.

The vertical distribution of PCBs was measured in a dated core from the continental shelf of the Korea Strait. The historical trend of deposition and fluxes of individual and total PCB agreed well with the dynamics of historical tendencies of PCBs consumption in South Korea up to mid-1980 and a peak of total PCBs of 810 ng kg−1 was recorded at the turn of the 1960s. A second period of deposition, when concentrations of total PCB up to 1007 ng kg−1 were reached, was evidenced between 1989 and 2010 after the ban of PCBs. Sediments deposited in 1990s–2010, and sediments deposited in the period of maximum PCBs use (1956–1983) are both enriched in the less chlorinated homologue groups (tri- and tetrachlorobyphenyls) and congener PCB 118. However, the ratio of dioxin-like PCBs, and dichloro- and hexachloro homologue groups disclosed compositional variations between the two time periods. Source analysis suggested that PCBs in the sediment record mainly originated from Kanechlor 300, Kanechlor 400 and Aroclor 1242 technical mixtures overlapped by secondary/unintentional combustion sources from regional steel making processes in the last decades (1990s-2010).

Network topology plays an important role in many network operations. However, it is very difficult to obtain the topology of public networks due to the lack of internal cooperation. Network tomography provides a powerful solution that can infer the network routing topology from end-to-end measurements. Existing solutions all assume that routes from a single source form a tree. However, with the rapid deployment of Software Defined Networking (SDN) and Network Function Virtualization (NFV), the routing paths in modern networks are becoming more complex. To address this problem, we propose a novel inference problem, called the weight inference problem, which infers the finest-granularity information from end-to-end measurements on general routing paths in general topologies. Our measurements are based on emulated multicast probes with a controllable “width”. We show that the problem has a unique solution when the multicast width is unconstrained; otherwise, we show that the problem can be treated as a sparse approximation problem, which allows us to apply variations of the pursuit algorithms. Simulations based on real network topologies show that our solution significantly outperforms a state-of-the-art network tomography algorithm, and increasing the width of multicast substantially improves the inference accuracy.

We introduce a new online learning framework where, at each trial, the learner is required to select a subset of actions from a given known action set. Each action is associated with an energy value, a reward and a cost. The sum of the energies of the actions selected cannot exceed a given energy budget. The goal is to maximise the cumulative profit, where the profit obtained on a single trial is defined as the difference between the maximum reward among the selected actions and the sum of their costs. Action energy values and the budget are known and fixed. All rewards and costs associated with each action change over time and are revealed at each trial only after the learner's selection of actions. Our framework encompasses several online learning problems where the environment changes over time; and the solution trades-off between minimising the costs and maximising the maximum reward of the selected subset of actions, while being constrained to an action energy budget. The algorithm that we propose is efficient and general in that it may be specialised to multiple natural online combinatorial problems., Published in AISTATS 2019

Mobile edge computing (MEC) is a promising technique for providing low-latency access to services at the network edge. The services are hosted at various types of edge nodes with both computation and communication capabilities. Due to the heterogeneity of edge node characteristics and user locations, the performance of MEC varies depending on where the service is hosted. In this paper, we consider such a heterogeneous MEC system, and focus on the problem of placing multiple services in the system to maximize the total reward. We show that the problem is NP-hard via reduction from the set cover problem, and propose a deterministic approximation algorithm to solve the problem, which has an approximation ratio that is not worse than $\left(1-e^{-1}\right)/4$. The proposed algorithm is based on two sub-routines that are suitable for small and arbitrarily sized services, respectively. The algorithm is designed using a novel way of partitioning each edge node into multiple slots, where each slot contains one service. The approximation guarantee is obtained via a specialization of the method of conditional expectations, which uses a randomized procedure as an intermediate step. In addition to theoretical guarantees, simulation results also show that the proposed algorithm outperforms other state-of-the-art approaches., Comment: Extended version of paper presented at IEEE INFOCOM 2019

The West Antarctic Ice Sheet (WAIS) Divide deep ice core WD2014 chronology, consisting of ice age, gas age, delta-age and uncertainties therein. The West Antarctic Ice Sheet Divide (WAIS Divide, WD) ice core is a newly drilled, high-accumulation deep ice core that provides Antarctic climate records of the past ~68 ka at unprecedented temporal resolution. The upper 2850 m (back to 31.2 ka BP; Sigl et al., 2015, Sigl et al., 2016) have been dated using annual-layer counting based on counting of annual layers observed in the chemical, dust and electrical conductivity records. The measurements were interpreted manually and with the aid of two automated methods. We validated the chronology by comparing of the cosmogenic isotope records of 10Be from WAIS Divide and 14C for IntCal13. We demonstrated that over the Holocene WD2014 was consistently accurate to better than 0.5% of the age. The chronology for the deep part of the core (below 2850m; 67.8-31.2 ka BP; Buizert et al., 2015) is based on stratigraphic matching to annual-layer-counted Greenland ice cores using globally well-mixed atmospheric methane. We calculate the WD gas age-ice age difference (Delta age) using a combination of firn densification modeling, ice-flow modeling, and a data set of d15N-N2, a proxy for past firn column thickness. The largest Delta age at WD occurs during the Last Glacial Maximum, and is 525 +/- 120 years. We synchronized the WD chronology to a linearly scaled version of the layer-counted Greenland Ice Core Chronology (GICC05), which brings the age of Dansgaard-Oeschger (DO) events into agreement with the U/Th absolutely dated Hulu Cave speleothem record.

Brain Computer Interfaces (BCIs) are interfaces that put the user in communication with an electronic device mainly through brain signals. BCIs may offer promising applications for impaired patient assistance by controlling computers and/or robotic devices based on the patient's intentions. Non-invasive BCIs are needed for users acceptance, but they have several issues, as signals generated by the brain are dampened by the skull. In this paper, we propose some practical usage examples of two non-invasive commercial BCIs, experimented in a set of active tasks to control robot's movements. Through experiments, the commercial BCIs showed initial promising results, even if other complementary inputs were also used, e.g. and electromyography bracelet and eye and clench movements.

Today, most information sources provide factual, objective knowledge, but they fail to capture personalized contextual knowledge which could be used to enrich the available factual data and contribute to their interpretation, in the context of the knowledge of the user who queries the system. This would require a knowledge framework which can accommodate both objective data and seman- tic enrichments that capture user provided knowledge associated to the factual data in the database. Unfortunately, most conventional DBMSs lack the flexibilities necessary (a) to prevent the data and metadata, evolve quickly with changing application requirements and (b) to capture user-provided and/or crowdsourced data and knowledge for more effective decision support. In this paper, we present CrowdSourced Semantic Enrichment (CroSSE) knowledge framework which al- lows traditional databases and semantic enrichment modules to coexist. CroSSE provides a novel Semantically Enriched SQL (SESQL) language to enrich SQL queries with information from a knowledge base containing semantic annotations. We describe CroSSE and SESQL with examples taken from our SmartGround EU project.

The Life Cycle Assessment (LCA) methodology is widely applied in several industrial sectors to evaluate the environmental performance of processes, products and services. Recently, several reports and studies have emphasized the importance of LCA in the field of engineered nanomaterials. However, to date only a few LCA studies on nanotechnology have been carried out, and fewer still have assessed aspects relating to ecotoxicity. This is mainly due to the lack of knowledge in relation on human and environmental exposure and effect of engineered nanoparticles (ENPs). This bottleneck is continued when performing Life Cycle Impact Assessment, where characterization models and consequently characterization factors (CFs) for ENPs are missing. This paper aims to provide the freshwater ecotoxicity CF for titanium dioxide nanoparticles (nano-TiO2). The USEtox™ model has been selected as a characterisation model. An adjusted multimedia fate model has been developed which accounts for nano-specific fate process descriptors (i.e. sedimentation, aggregation with suspended particle matter, etc.) to estimate the fate of nano-TiO2 in freshwater. A literature survey of toxicity tests performed on freshwater organism representative of multiple trophic levels was conducted, including algae, crustaceans and fish in order to collect relevant EC50 values. Then, the toxic effect of nano-TiO2 was computed on the basis of the HC50 value. Thus, following the principle of USEtox™ model and accounting for nano-specific descriptors a CF for the toxic impact of freshwater ecotoxicity of 0.28 PAF day m3 kg− 1 is proposed.

Currently, a noncomprehensive understanding of the physicochemical properties of carbon-based nanomaterial (CBNs), which may affect toxic effects, is still observable. In this study, an exploratory systematic investigation into the key physicochemical properties of multiwall carbon nanotube (MWCNT), single-wall carbon nanotube (SWCNT), and C60-fullerene on their ecotoxicity has been undertaken. We undertook an extensive survey of the literature pertaining to the ecotoxicity of organism representative of the trophic level of algae, crustaceans, and fish. Based on this, a set of data reporting both the physicochemical properties of carbon-based nanomaterial and the observed toxic effect has been established. The relationship between physicochemical properties and observed toxic effect was investigated based on various statistical approaches. Specifically, analysis of variance by one-way ANOVA was used to assess the effect of categorical properties (use of a dispersant or treatments in the test medium, type of carbon-based nanomaterial, i.e., SWCNT, MWCNT, C60-fullerene, functionalization), while multiple regression analysis was used to assess the effect of quantitative properties (i.e., diameter length of nanotubes, secondary size) on the toxicity values. The here described investigations revealed significant relationships among the physicochemical properties and observed toxic effects. The research was mainly affected by the low availability of data and also by the low variability of the studies collected. Overall, our results demonstrate that the here proposed and applied approach could have a major role in identifying the physicochemical properties of relevance for the toxicity of nanomaterial. However, the future success of the approach would require that the ENMs and the experimental conditions used in the toxicity studies are fully characterized.

Pharmaceuticals are widespread emerging contaminants and, like all pollutants, are present in combination with others in the ecosystems. The aim of the present work was to evaluate the toxic response of the crustacean Daphnia magna exposed to individual and combined pharmaceuticals. Fluoxetine, a selective serotonin re-uptake inhibitor widely prescribed as antidepressant, and propranolol, a non-selective β-adrenergic receptor-blocking agent used to treat hypertension, were tested. Several experimental trials of an acute immobilization test and a chronic reproduction test were performed. Single chemicals were first tested separately. Toxicity of binary mixtures was then assessed using a fixed ratio experimental design. Five concentrations and 5 percentages of each substance in the mixture (0, 25, 50, 75, and 100%) were tested. The MIXTOX model was applied to analyze the experimental results. This tool is a stepwise statistical procedure that evaluates if and how observed data deviate from a reference model, either concentration addition (CA) or independent action (IA), and provides significance testing for synergism, antagonism, or more complex interactions. Acute EC50 values ranged from 6.4 to 7.8 mg/L for propranolol and from 6.4 to 9.1 mg/L for fluoxetine. Chronic EC50 values ranged from 0.59 to 1.00 mg/L for propranolol and from 0.23 to 0.24 mg/L for fluoxetine. Results showed a significant antagonism between chemicals in both the acute and the chronic mixture tests when CA was adopted as the reference model, while absence of interactive effects when IA was used.

Fatty acid amide hydrolase (FAAH) is a membrane anchored serine hydrolase that has a principle role in the metabolism of the endogenous cannabinoid anandamide. Docking studies using representative FAAH crystal structures revealed that compounds containing a novel piperidinyl thiazole isoxazoline core fit within the ligand binding domains. New potential FAAH inhibitors were designed and synthesized incorporating urea, carbamate, alkyldione and thiourea reactive centers as potential pharmacophores. A small library of candidate compounds (75) was then screened against human FAAH leading to the identification of new carbamate and urea based inhibitors ( K i = pM and nM, respectively). Representative carbamate and urea based chemotypes displayed slow, time dependent inhibition kinetics leading to enzyme inactivation which was slowly reversible. However, evidence indicated that features of the mechanism of inactivation differ between the two pharmacophore types. Selected compounds were also evaluated for analgesic activity in the mouse-tail flick test.

The effect of the administration of beneficial lactic acid bacteria (LAB): Lactococcus lactis CRL 1584, L. lactis CRL 1827, Enterococcus gallinarum CRL 1826 and combined CRL 1584+ CRL 1826 on the development of bullfrog embryos from the hatching stage until 31 days (tadpole) was evaluated. In vitro assays indicated that 103, 106 and 109 CFU mL−1 single LAB strains remained viable until 24 h in 10% Ringer. Around 109 CFU mL−1 LAB (individually and combined) were used in an experimental design built to evaluate their effect when administrated at different intervals (three 7-day cycles with 5-day rest periods in between) to embryos until day 31. The highest potentially beneficial population (LAB) numbers were detected in the LAB-treated groups. All the LAB-treatments increased it and were significantly higher than the controls. Although the highest, potentially pathogenic, population (Red-Leg Syndrome-RLS-related pathogens) numbers were detected in the control and the lowest in the CRL 1584+ CRL 1826-treated group, they did not differ significantly. Stereoscopic studies showed no malformations in any LAB-treated group and all the specimens reached the same stage of their biological cycle with a survival >94%. The histological structure of target organs for RLS-associated pathogens (intestine and skin) and stomach was not affected and the spleen was developed. Only the LAB-treated groups showed microorganisms associated with the intestinal mucus, without inflammatory reaction in the lamina propria. This is the first report on the evaluation of the absence of adverse effects after LAB administration to bullfrog embryos using indigenous microorganisms. Fil: Pasteris, Sergio Enrique. Consejo Nacional de Investigaciones Científicas y Técnicas. Centro Científico Tecnológico Conicet - Tucumán. Instituto Superior de Investigaciones Biológicas. Universidad Nacional de Tucumán. Instituto Superior de Investigaciones Biológicas; Argentina Fil: Montel Mendoza, María Gabriela. Consejo Nacional de Investigaciones Científicas y Técnicas. Centro Científico Tecnológico Conicet - Tucumán. Instituto Superior de Investigaciones Biológicas. Universidad Nacional de Tucumán. Instituto Superior de Investigaciones Biológicas; Argentina Fil: Llanos, Ricardo J.. Consejo Nacional de Investigaciones Científicas y Técnicas. Centro Científico Tecnológico Conicet - Tucumán. Instituto Superior de Investigaciones Biológicas. Universidad Nacional de Tucumán. Instituto Superior de Investigaciones Biológicas; Argentina Fil: Pucci Alcaide, Franco J.. Fundación Miguel Lillo; Argentina Fil: Nader, Maria Elena Fatima. Consejo Nacional de Investigaciones Científicas y Técnicas. Centro Científico Tecnológico Conicet - Tucuman. Centro de Referencia Para Lactobacilos; Argentina

Phosphate is commonly added to drinking water to inhibit lead release from lead service lines and lead-containing materials in premise plumbing. Phosphate addition promotes the formation of lead phosphate particles, and their aggregation behaviors may affect their transport in pipes. Here, lead phosphate formation and aggregation were studied under varied aqueous conditions typical of water supply systems. Under high aqueous PO4/Pb molar ratios (>1), phosphate adsorption made the particles more negatively charged. Therefore, enhanced stability of lead phosphate particles was observed, suggesting that although addition of excess phosphate can lower the dissolved lead concentrations in tap water, it may increase concentrations of particulate lead. Adsorption of divalent cations (Ca2+ and Mg2+) onto lead phosphate particles neutralized their negative surface charges and promoted their aggregation at pH 7, indicating that phosphate addition for lead immobilization may be more efficient in harder waters. The p...

In this paper, we discuss the problem of distributed learning for coalition operations. We consider a scenario where different coalition forces are running learning systems independently but want to merge the insights obtained from all the learning systems to share knowledge and use a single model combining all of their individual models. We consider the challenges involved in such fusion of models, and propose an algorithm that can find the right fused model in an efficient manner.

The outermost enamel of the human tooth and the rostrum of the whale Mesoplodon densirostris are two highly mineralized tissues that contain over 95 wt.% mineral, i.e., bioapatite. However, the same mineral type (carbonated hydroxylapatite) does not yield the same material properties, as revealed by Raman spectroscopy, scanning electron microscopy, electron microprobe analysis, and synchrotron X-ray diffraction analysis. Overall, the outermost enamel of a tooth has more homogeneous physical and chemical features than the rostrum. Chemical comparison of rostrum and enamel shows bioapatite in the rostrum to be enriched in Na, Mg, CO3, and S, whereas the outermost enamel shows only a slightly enriched Cl concentration. Morphologically, mineral rods (at tens of μm scale), crystallites and prisms (at μm and sub-μm scale), and platelets (at tens of nm scale) all demonstrate less organized texture in the rostrum than in enamel. Such contrasts between two mineralized tissues suggest distinct pathways of biomineralization, e.g., the nature of the equilibrium between mineral and body fluid. This study illustrates the remarkable flexibility of the apatite mineral structure to match its chemical and physical properties to specific biological needs within the same animal or between species.