Efficient energy-based embedding models for link prediction in knowledge graphs

We focus on the problem of link prediction in Knowledge Graphs, with the goal of discovering new facts. To this purpose, Energy-Based Models for Knowledge Graphs that embed entities and relations in continuous vector spaces have been largely used. The main limitation in their applicability lies in the parameter learning phase, which may require a large amount of time for converging to optimal solutions. In this article, we first propose an unified view on different Energy-Based Embedding Models. Hence, for improving the model training phase, we propose the adoption of adaptive learning rates. We show that, by adopting adaptive learning rates during training, we can improve the efficiency of the parameter learning process by an order of magnitude, while leading to more accurate link prediction models in a significantly lower number of iterations. We extensively evaluate the proposed learning procedure on a variety of new models: our result show a significant improvement over state-of-the-art link prediction methods on two large Knowledge Graphs, namely WordNet and Freebase.     

Adsorption of pharmaceuticals from aqueous solutions on synthetic zeolites

The objectives of this study were to measure the sorption capacity of a dilute solution of drugs in water on hydrophobic zeolites and to quantify aspects of their adsorption mechanisms for potential use in wastewater remediation. The removal of three drugs, erythromycin (ERY), carbamazepine (CBZ) and levofloxacin (FLX) from water using three organophilic zeolites (Y, mordenite (MOR), ZSM-5) was investigated. Adsorption isotherms show that these three drugs are adsorbed in remarkable amounts by Y. These data were confirmed by thermogravimetric analysis. The presence of these drugs inside the Y cage was revealed by unit cell parameter variations and structural deformations obtained by X-ray structure analyses carried out using the Rietveld method on zeolite after adsorption. Our results indicate that the adsorption properties of zeolitic materials do not only depend on micropore size and that zeolite shape selectivity also depends on structural features which are difficult to determine “a priori”. In order to test the ability of organophilic zeolites in removing drugs from wastewater, the adsorption of ERY, FLX and CBZ on zeolite Y from water samples collected at the outlet of a wastewater treatment plant located in Northern Italy was determined. These drugs were almost completely adsorbed by Y zeolite, confirming that this material is suitable for removing drugs from wastewater.     

SafeNet: A methodology for integrating general-purpose unsafe devices in safe-robot rehabilitation systems

Robot-assisted neurorehabilitation often involves networked systems of sensors (“sensory rooms”) and powerful devices in physical interaction with weak users. Safety is unquestionably a primary concern. Some lightweight robot platforms and devices designed on purpose include safety properties using redundant sensors or intrinsic safety design (e.g. compliance and backdrivability, limited exchange of energy). Nonetheless, the entire “sensory room” shall be required to be fail-safe and safely monitored as a system at large. Yet, sensor capabilities and control algorithms used in functional therapies require, in general, frequent updates or re-configurations, making a safety-grade release of such devices hardly sustainable in cost-effectiveness and development time. As such, promising integrated platforms for human-in-the-loop therapies could not find clinical application and manufacturing support because of lacking in the maintenance of global fail-safe properties.     

Probabilistic topic models for sequence data

Probabilistic topic models are widely used in different contexts to uncover the hidden structure in large text corpora. One of the main (and perhaps strong) assumption of these models is that generative process follows a bag-of-words assumption, i.e. each token is independent from the previous one. We extend the popular Latent Dirichlet Allocation model by exploiting three different conditional Markovian assumptions: (i) the token generation depends on the current topic and on the previous token; (ii) the topic associated with each observation depends on topic associated with the previous one; (iii) the token generation depends on the current and previous topic. For each of these modeling assumptions we present a Gibbs Sampling procedure for parameter estimation. Experimental evaluation over real-word data shows the performance advantages, in terms of recall and precision, of the sequence-modeling approaches.     

Two fault‐tolerant control problems for multiple‐integrator networks

The paper considers a network of agents with multiple‐integrator internal dynamics, which share partial information on their states according to an arbitrary topology. For this system, two control problems are addressed and solved. The first consists in assigning the dominant closed‐loop poles. The second consists in achieving a specified consensus with arbitrary dominant dynamics. In both cases, the regulator is required to be decentralized, and the controlled network has to result tolerant with respect to faults in the communication apparatuses of the agents. Copyright © 2015 John Wiley & Sons, Ltd.     

Novel poly(3‐hydroxybutyrate) composite films containing bioactive glass nanoparticles for wound healing applications

Bioactive glass is considered an ideal material for haemostasis as it releases Ca²⁺ ions upon hydration, which is required to support thrombosis. In this study the effects of the presence of nanoscaled bioactive glass (n‐BG) in poly(3‐hydroxybutyrate) (P(3HB)) microsphere films on the structural properties, thermal properties and biocompatibility of the films were studied. The n‐BG with a high surface area was also tested for its in vitro haemostatic efficacy and was found to be able to successfully reduce clot detection time. In an effort to study the effect of the roughness induced by the formation of hydroxyapatite on cellular functions such as cell adhesion, cell mobility and cell differentiation, the composite films were immersed in simulated body fluid for periods of 1, 3 and 7 days. From scanning electron microscopy images, the surface of the P(3HB)/n‐BG composite microsphere films appeared fairly uniform and smooth on day 1; however on day 3 and day 7 a rough and uneven surface was observed. The presence of hydroxyapatite on the composite microsphere films on day 3 and day 7 influenced the surface roughness of the films. However, when the P(3HB)/n‐BG composite microsphere films with enhanced surface roughness were tested for biocompatibility, reduced amounts of protein adsorption and cell adhesion were observed. This study thus revealed that there is an optimal surface roughness for the P(3HB) microsphere films for increased cell adhesion, beyond which it could be deleterious for cell adhesion and differentiation. © 2016 Society of Chemical Industry     

Optical and electrical characterization of a gold nanoparticle dispersion in a chiral liquid crystal matrix

We report on the effect of gold nanoparticle (Au NP) dispersion in a chiral nematic liquid crystal (LC). Polarized optical microscopy and X-ray diffraction measurements evidence the insurgence of an order change in the LC host. Moreover, a comparative analysis based on dielectric and voltammetric spectroscopies performed on pure LC and on Au NP-doped LC shows that Au NP’s presence besides affecting LC order influences its electric properties: ion conductivity results importantly reduced, and beyond a threshold value of the applied field electrophoresis phenomena are induced.