Gianni Fenu

An approach to a Cloud Computing network

ldquoCloud Computingrdquo is becoming increasingly relevant, as it will enable companies involved in spreading this technology to open the doors to Web 3.0. In this work the basic features of cloud computing are presented and compared with those of the original technology: Grid Computing. The new categories of services introduced will slowly replace many types of computational resources currently used. In this perspective, grid computing, the basic element for the large scale supply of cloud services, will play a fundamental role in defining how those services will be provided. The paper describes the concept of computational resources outsourcing, referred to computational grids and a real application. This work utilises the results by the Cybersar Project managed by the COSMOLAB Consortium (Italy).

The urinary metabolomics profile of an Italian autistic children population and their unaffected siblings

Abstract Objective: A supervised multivariate model to classify the metabolome alterations between autistic spectrum disorders (ASD) patients and controls, siblings of autistic patients, has been realized and used to realize a network model of the ASD patients’ metabolome. Methods: In our experiment we propose a quantification of urinary metabolites with the Mass Spectroscopy technique couple to Gas Chromatography. A multivariate model has been used to extrapolate the variables of importance for a network model of interaction between metabolites. In this way we are able to propose a network-based approach to ASD description. Results: Children with autistic disease composing our studied population showed elevated concentration of several organic acids and sugars. Interactions among diet, intestinal flora and genes may explain such findings. Among them, the 3-(3-hydroxyphenyl)-3-hydroxypropanoic acid has been previously described as altered in autistic subjects. Other metabolites increased are 3,4-dihydroxybutyric acid, glycolic acid and glycine, cis-aconitic acid; phenylalanine, tyrosine, p-hydroxyphenylacetic acid, and homovanillic acid are all involved in the tyrosine pathway leading to neurotransmitter cathecolamine. Conclusion: GC-MS-based metabolomic analysis of the urinary metabolome suggests to have the required sensitivity and specificity to gain insight into ASD phenotypes and aid a personalized network-based medicine approach.

GestIT: a declarative and compositional framework for multiplatform gesture definition

Gestural interfaces allow complex manipulative interactions that are hardly manageable using traditional event handlers. Indeed, such kind of interaction has longer duration in time than that carried out in form-based user interfaces, and often it is important to provide users with intermediate feedback during the gesture performance. Therefore, the gesture specification code is a mixture of the recognition logic and the feedback definition. This makes it difficult 1) to write maintainable code and 2) reuse the gesture definition in different applications. To overcome these kinds of limitations, the research community has considered declarative approaches for the specification of gesture temporal evolution. In this paper, we discuss the creation of gestural interfaces using GestIT, a framework that allows declarative and compositional definition of gestures for different recognition platforms (e.g. multitouch and full-body), through a set of examples and the comparison with existing approaches.

Urinary metabolomics (GC-MS) reveals that low and high birth weight infants share elevated inositol concentrations at birth.

Abstract Objective: Metabolomics is a new “omics” platform aimed at high-throughput identification, quantification and characterization of small-molecule metabolites. The metabolomics approach has been successfully applied to the classification different physiological states and identification of perturbed biochemical pathways. The purpose of the current investigation is the application of metabolomics to explore biological mechanisms which may lead to the onset of metabolic syndrome in adulthood. Methods: We evaluated differences in metabolites in the urine collected within 12 h from 23 infants with IUGR (IntraUterine Growth Restriction), or LGA (Large for Gestational Age), compared to control infants (10 patients defined AGA: Appropriate for Gestational Age). Urinary metabolites were quantified by GC-MS and used to highlight similarities between the two metabolic diseases and identify metabolic markers for their predisposition. Quantified metabolites were analyzed using a multivariate statistics coupled with receiver operator characteristic curve (ROC) analysis of identified biomarkers. Results: Urinary myo-inositol was the most important discriminant between LGA + IUGR and control infants, and displayed an area under the ROC curve = 1. Conclusion: We postulate that the increase in plasma and consequently urinary inositol may constitute a marker of altered glucose metabolism during fetal development in both IUGR and LGA newborns.

White Paper on Industrial Applications of Computer Vision and Pattern Recognition

The paper provides a summary of the contributions to the industrial session at ICIAP2013, describing a few practical applications of Video Analy- sis, in the Surveillance and Security field. The session has been organized to stimulate an open discussion within the scientific community of CVPR on new emerging research areas which deserve particular attention, and may contribute to the improvement of industrial applications in the near future.

A Cloud Computing Based Real Time Financial System

The “Cloud Computing” is becoming an increasingly popular term. The new “XaaS” category of services introduced will slowly replace many types of computational resources currently used. In this perspective, grid computing, the basic element for the large scale supply of cloud services, will play a fundamental role in defining how those services will be provided. This paper is concerned with the study and preliminary design of a real time financial system based on cloud computing technologies that enable macroeconomic analysis and forecasts of the financial markets and their instruments. Cloud and Grid paradigms can generate different added values which are examined in detail in the paper. This work utilises the results obtained during the Cybersar Project managed by the COSMOLAB Consortium (Italy). The system analyzed and described herein will be implemented in the Cybersar Computational Grid by autumn 2009.

RFID- based supply chain traceability system

In this paper we propose a real solution for gathering information throughout the entire pig meat supply chain. The architecture consists of a a complex identification system based on RFID tags that transmits data to a distributed database during all phases of the production process. The specific work environment required identifying a suitable technology for implementation in the supply chain and the best possible organization. The aim of this work is to keep track of all the information generated during meat processing, not only for traceability purposes but chiefly for enhancing and optimizing production. All information generated by the traceability system will be collected in a central database accessible by end users thtough a public dedicated web interface.

Strategies to Carry and Forward Packets in VANET

The aim of this paper is to find the best strategies to carry and forward packets within VANETs that follows a Delay Tolerant Network. In this environment nodes are affected by intermittent connectivity and topology constantly changes. When no route is available and the link failure percentage is high, the data must be physically transported by vehicles to destination. Results show how, using vehicles cooperation and several carry and forward mechanisms with different deliver priorities, is possible to improve performance for free in terms of data delivery.

Network-adaptive management of computation energy in wireless sensor networks

Todays sensor nodes can be equipped with powerful microcontrollers to address the increasing need of real-time processing of sensed data. For instance, body sensor networks for gesture recognition require filtering of acceleration values at line rate. This requirement imposes a paradigm shift with regard to more traditional sensor networks characterized by low activity duty cycles. Therefore, energy conservation strategies applied to wireless sensor nodes to increase their lifetime must take into account computation power rather than focusing only on communication power. In this paper we present a novel approach which aims at exploiting the knowledge of network status to optimize the power consumption of the node microcontroller. The proposed approach is tested in various network conditions, both synthetic and realistic, in the context of IEEE 802.15.4 standard. Experimental results demonstrate that the proposed approach allows to achieve power savings of up to 70% with minimum performance penalty.

Using neural word embeddings to model user behavior and detect user segments

Modeling user behavior to detect segments of users to target and to whom address ads (behavioral targeting) is a problem widely-studied in the literature. Various sources of data are mined and modeled in order to detect these segments, such as the queries issued by the users. In this paper we first show the need for a user segmentation system to employ reliable user preferences, since nearly half of the times users reformulate their queries in order to satisfy their information need. Then we propose a method that analyzes the description of the items positively evaluated by the users and extracts a vector representation of the words in these descriptions (word embeddings). Since it is widely-known that users tend to choose items of the same categories, our approach is designed to avoid the so-called preference stability, which would associate the users to trivial segments. Moreover, we make sure that the interpretability of the generated segments is a characteristic offered to the advertisers who will use them. We performed different sets of experiments on a large real-world dataset, which validated our approach and showed its capability to produce effective segments.