Semi-supervised learning with concept drift using particle dynamics applied to network intrusion detection data

Concept drift, which refers to non stationary learning problems over time, has increasing importance in machine learning and data mining. Many concept drift applications require fast response, which means an algorithm must always be (re)trained with the latest available data. But the process of data labeling is usually expensive and/or time consuming when compared to acquisition of unlabeled data, thus usually only a small fraction of the incoming data may be effectively labeled. Semi-supervised learning methods may help in this scenario, as they use both labeled and unlabeled data in the training process. However, most of them are based on assumptions that the data is static. Therefore, semi-supervised learning with concept drifts is still an open challenging task in machine learning. Recently, a particle competition and cooperation approach has been developed to realize graph-based semi-supervised learning from static data. We have extend that approach to handle data streams and concept drift. The result is a passive algorithm which uses a single classifier approach, naturally adapted to concept changes without any explicit drift detection mechanism. It has built-in mechanisms that provide a natural way of learning from new data, gradually "forgetting" older knowledge as older data items are no longer useful for the classification of newer data items. The proposed algorithm is applied to the KDD Cup 1999 Data of network intrusion, showing its effectiveness.

Semantic integration of gene expression analysis tools and data sources using software connectors

Abstract Background The study and analysis of gene expression measurements is the primary focus of functional genomics. Once expression data is available, biologists are faced with the task of extracting (new) knowledge associated to the underlying biological phenomenon. Most often, in order to perform this task, biologists execute a number of analysis activities on the available gene expression dataset rather than a single analysis activity. The integration of heteregeneous tools and data sources to create an integrated analysis environment represents a challenging and error-prone task. Semantic integration enables the assignment of unambiguous meanings to data shared among different applications in an integrated environment, allowing the exchange of data in a semantically consistent and meaningful way. This work aims at developing an ontology-based methodology for the semantic integration of gene expression analysis tools and data sources. The proposed methodology relies on software connectors to support not only the access to heterogeneous data sources but also the definition of transformation rules on exchanged data. Results We have studied the different challenges involved in the integration of computer systems and the role software connectors play in this task. We have also studied a number of gene expression technologies, analysis tools and related ontologies in order to devise basic integration scenarios and propose a reference ontology for the gene expression domain. Then, we have defined a number of activities and associated guidelines to prescribe how the development of connectors should be carried out. Finally, we have applied the proposed methodology in the construction of three different integration scenarios involving the use of different tools for the analysis of different types of gene expression data. Conclusions The proposed methodology facilitates the development of connectors capable of semantically integrating different gene expression analysis tools and data sources. The methodology can be used in the development of connectors supporting both simple and nontrivial processing requirements, thus assuring accurate data exchange and information interpretation from exchanged data.

Ecucational data mining (EDM) para la determinación de comportamientos en estudiantes de ingeniería en la modalidad virtual de UDE

[ES]En este artículo se describe la experiencia de la aplicación de técnicas de EDM (clustering) a un curso disponible en la plataforma Ude@ de la Universidad de Antioquia. El objetivo es clasificar los patrones de interacción de los estudiantes a partir de la información almacenada en la base de datos de la plataforma Moodle. Para ello, se generan informes sobre el uso de los recursos y la autoevaluación que permiten analizar el comportamiento y los patrones de navegación de los estudiantes durante el uso del LMS (Learning Management System).

Memory hierarchy and data communication in heterogeneous reconfigurable SoCs

The miniaturization race in the hardware industry aiming at continuous increasing of transistor density on a die does not bring respective application performance improvements any more. One of the most promising alternatives is to exploit a heterogeneous nature of common applications in hardware. Supported by reconfigurable computation, which has already proved its efficiency in accelerating data intensive applications, this concept promises a breakthrough in contemporary technology development. Memory organization in such heterogeneous reconfigurable architectures becomes very critical. Two primary aspects introduce a sophisticated trade-off. On the one hand, a memory subsystem should provide well organized distributed data structure and guarantee the required data bandwidth. On the other hand, it should hide the heterogeneous hardware structure from the end-user, in order to support feasible high-level programmability of the system. This thesis work explores the heterogeneous reconfigurable hardware architectures and presents possible solutions to cope the problem of memory organization and data structure. By the example of the MORPHEUS heterogeneous platform, the discussion follows the complete design cycle, starting from decision making and justification, until hardware realization. Particular emphasis is made on the methods to support high system performance, meet application requirements, and provide a user-friendly programmer interface. As a result, the research introduces a complete heterogeneous platform enhanced with a hierarchical memory organization, which copes with its task by means of separating computation from communication, providing reconfigurable engines with computation and configuration data, and unification of heterogeneous computational devices using local storage buffers. It is distinguished from the related solutions by distributed data-flow organization, specifically engineered mechanisms to operate with data on local domains, particular communication infrastructure based on Network-on-Chip, and thorough methods to prevent computation and communication stalls. In addition, a novel advanced technique to accelerate memory access was developed and implemented.

Clinical Data Mining: Traumatic Brain Injury

Il trauma cranico é tra le piú importanti patologie traumatiche. Ogni anno 250 pazienti ogni 100.000 abitanti vengono ricoverati in Italia per un trauma cranico. La mortalitá é di circa 17 casi per 100.000 abitanti per anno. L’Italia si trova in piena “media” Europea considerando l’incidenza media in Europa di 232 casi per 100.000 abitanti ed una mortalitá di 15 casi per 100.000 abitanti. Degli studi hanno indicato come una terapia anticoagulante é uno dei principali fattori di rischio di evolutiviá di una lesione emorragica. Al contrario della terapia anticoagulante, il rischio emorragico correlato ad una terapia antiaggregante é a tutt’oggi ancora in fase di verifica. Il problema risulta rilevante in particolare nella popolazione occidentale in quanto l’impiego degli antiaggreganti é progressivamente sempre piú diffuso. Questo per la politica di prevenzione sostenuta dalle linee guida nazionali e internazionali in termini di prevenzione del rischio cardiovascolare, in particolare nelle fasce di popolazione di etá piú avanzata. Per la prima volta, é stato dimostrato all’ospedale di Forlí[1], su una casistica sufficientemente ampia, che la terapia cronica con antiaggreganti, per la preven- zione del rischio cardiovascolare, puó rivelarsi un significativo fattore di rischio di complicanze emorragiche in un soggetto con trauma cranico, anche di grado lieve. L’ospedale per approfondire e convalidare i risultati della ricerca ha condotto, nell’anno 2009, una nuova indagine. La nuova indagine ha coinvolto oltre l’ospedale di Forlí altri trentuno centri ospedalieri italiani. Questo lavoro di ricerca vuole, insieme ai ricercatori dell’ospedale di Forlí, verificare: “se una terapia con antiaggreganti influenzi l’evolutivitá, in senso peggiorativo, di una lesione emorragica conseguente a trauma cranico lieve - moderato - severo in un soggetto adulto”, grazie ai dati raccolti dai centri ospedalieri nel 2009. Il documento é strutturato in due parti. La prima parte piú teorica, vuole fissare i concetti chiave riguardanti il contesto della ricerca e la metodologia usata per analizzare i dati. Mentre, la seconda parte piú pratica, vuole illustrare il lavoro fatto per rispondere al quesito della ricerca. La prima parte é composta da due capitoli, che sono: • Il capitolo 1: dove sono descritti i seguenti concetti: cos’é un trauma cra- nico, cos’é un farmaco di tipo anticoagulante e cos’é un farmaco di tipo antiaggregante; • Il capitolo 2: dove é descritto cos’é il Data Mining e quali tecniche sono state usate per analizzare i dati. La seconda parte é composta da quattro capitoli, che sono: • Il capitolo 3: dove sono state descritte: la struttura dei dati raccolti dai trentadue centri ospedalieri, la fase di pre-processing e trasformazione dei dati. Inoltre in questo capitolo sono descritti anche gli strumenti utilizzati per analizzare i dati; • Il capitolo 4: dove é stato descritto come é stata eseguita l’analisi esplorativa dei dati. • Il capitolo 5: dove sono descritte le analisi svolte sui dati e soprattutto i risultati che le analisi, grazie alle tecniche di Data Mining, hanno prodotto per rispondere al quesito della ricerca; • Il capitolo 6: dove sono descritte le conclusioni della ricerca. Per una maggiore comprensione del lavoro sono state aggiunte due appendici. La prima tratta del software per data mining Weka, utilizzato per effettuare le analisi. Mentre, la seconda tratta dell’implementazione dei metodi per la creazione degli alberi decisionali.

Procedures of quality control and data analysis of multi-site ground-based observations for the absolute flux calibration of Gaia

The Gaia space mission is a major project for the European astronomical community. As challenging as it is, the processing and analysis of the huge data-flow incoming from Gaia is the subject of thorough study and preparatory work by the DPAC (Data Processing and Analysis Consortium), in charge of all aspects of the Gaia data reduction. This PhD Thesis was carried out in the framework of the DPAC, within the team based in Bologna. The task of the Bologna team is to define the calibration model and to build a grid of spectro-photometric standard stars (SPSS) suitable for the absolute flux calibration of the Gaia G-band photometry and the BP/RP spectrophotometry. Such a flux calibration can be performed by repeatedly observing each SPSS during the life-time of the Gaia mission and by comparing the observed Gaia spectra to the spectra obtained by our ground-based observations. Due to both the different observing sites involved and the huge amount of frames expected (≃100000), it is essential to maintain the maximum homogeneity in data quality, acquisition and treatment, and a particular care has to be used to test the capabilities of each telescope/instrument combination (through the “instrument familiarization plan”), to devise methods to keep under control, and eventually to correct for, the typical instrumental effects that can affect the high precision required for the Gaia SPSS grid (a few % with respect to Vega). I contributed to the ground-based survey of Gaia SPSS in many respects: with the observations, the instrument familiarization plan, the data reduction and analysis activities (both photometry and spectroscopy), and to the maintenance of the data archives. However, the field I was personally responsible for was photometry and in particular relative photometry for the production of short-term light curves. In this context I defined and tested a semi-automated pipeline which allows for the pre-reduction of imaging SPSS data and the production of aperture photometry catalogues ready to be used for further analysis. A series of semi-automated quality control criteria are included in the pipeline at various levels, from pre-reduction, to aperture photometry, to light curves production and analysis.

Analisi di dati citofluorimetrici con tecniche di Data Mining

Il citofluorimetro è uno strumento impiegato in biologia genetica per analizzare dei campioni cellulari: esso, analizza individualmente le cellule contenute in un campione ed estrae, per ciascuna cellula, una serie di proprietà fisiche, feature, che la descrivono. L’obiettivo di questo lavoro è mettere a punto una metodologia integrata che utilizzi tali informazioni modellando, automatizzando ed estendendo alcune procedure che vengono eseguite oggi manualmente dagli esperti del dominio nell’analisi di alcuni parametri dell’eiaculato. Questo richiede lo sviluppo di tecniche biochimiche per la marcatura delle cellule e tecniche informatiche per analizzare il dato. Il primo passo prevede la realizzazione di un classificatore che, sulla base delle feature delle cellule, classifichi e quindi consenta di isolare le cellule di interesse per un particolare esame. Il secondo prevede l'analisi delle cellule di interesse, estraendo delle feature aggregate che possono essere indicatrici di certe patologie. Il requisito è la generazione di un report esplicativo che illustri, nella maniera più opportuna, le conclusioni raggiunte e che possa fungere da sistema di supporto alle decisioni del medico/biologo.

Supernova neutrinos in AMANDA and IceCube - Monte Carlo development and data analysis

Supernovae are among the most energetic events occurring in the universe and are so far the only verified extrasolar source of neutrinos. As the explosion mechanism is still not well understood, recording a burst of neutrinos from such a stellar explosion would be an important benchmark for particle physics as well as for the core collapse models. The neutrino telescope IceCube is located at the Geographic South Pole and monitors the antarctic glacier for Cherenkov photons. Even though it was conceived for the detection of high energy neutrinos, it is capable of identifying a burst of low energy neutrinos ejected from a supernova in the Milky Way by exploiting the low photomultiplier noise in the antarctic ice and extracting a collective rate increase. A signal Monte Carlo specifically developed for water Cherenkov telescopes is presented. With its help, we will investigate how well IceCube can distinguish between core collapse models and oscillation scenarios. In the second part, nine years of data taken with the IceCube precursor AMANDA will be analyzed. Intensive data cleaning methods will be presented along with a background simulation. From the result, an upper limit on the expected occurrence of supernovae within the Milky Way will be determined.

Data management and data analysis in the large European projects GEHA (GEnetics of Healthy Aging) and NU-AGE (NUtrition and AGEing): a bioinformatic approach

The aging process is characterized by the progressive fitness decline experienced at all the levels of physiological organization, from single molecules up to the whole organism. Studies confirmed inflammaging, a chronic low-level inflammation, as a deeply intertwined partner of the aging process, which may provide the “common soil” upon which age-related diseases develop and flourish. Thus, albeit inflammation per se represents a physiological process, it can rapidly become detrimental if it goes out of control causing an excess of local and systemic inflammatory response, a striking risk factor for the elderly population. Developing interventions to counteract the establishment of this state is thus a top priority. Diet, among other factors, represents a good candidate to regulate inflammation. Building on top of this consideration, the EU project NU-AGE is now trying to assess if a Mediterranean diet, fortified for the elderly population needs, may help in modulating inflammaging. To do so, NU-AGE enrolled a total of 1250 subjects, half of which followed a 1-year long diet, and characterized them by mean of the most advanced –omics and non –omics analyses. The aim of this thesis was the development of a solid data management pipeline able to efficiently cope with the results of these assays, which are now flowing inside a centralized database, ready to be used to test the most disparate scientific hypotheses. At the same time, the work hereby described encompasses the data analysis of the GEHA project, which was focused on identifying the genetic determinants of longevity, with a particular focus on developing and applying a method for detecting epistatic interactions in human mtDNA. Eventually, in an effort to propel the adoption of NGS technologies in everyday pipeline, we developed a NGS variant calling pipeline devoted to solve all the sequencing-related issues of the mtDNA.

Learning Methods and Algorithms for Semantic Text Classification across Multiple Domains

Information is nowadays a key resource: machine learning and data mining techniques have been developed to extract high-level information from great amounts of data. As most data comes in form of unstructured text in natural languages, research on text mining is currently very active and dealing with practical problems. Among these, text categorization deals with the automatic organization of large quantities of documents in priorly defined taxonomies of topic categories, possibly arranged in large hierarchies. In commonly proposed machine learning approaches, classifiers are automatically trained from pre-labeled documents: they can perform very accurate classification, but often require a consistent training set and notable computational effort. Methods for cross-domain text categorization have been proposed, allowing to leverage a set of labeled documents of one domain to classify those of another one. Most methods use advanced statistical techniques, usually involving tuning of parameters. A first contribution presented here is a method based on nearest centroid classification, where profiles of categories are generated from the known domain and then iteratively adapted to the unknown one. Despite being conceptually simple and having easily tuned parameters, this method achieves state-of-the-art accuracy in most benchmark datasets with fast running times. A second, deeper contribution involves the design of a domain-independent model to distinguish the degree and type of relatedness between arbitrary documents and topics, inferred from the different types of semantic relationships between respective representative words, identified by specific search algorithms. The application of this model is tested on both flat and hierarchical text categorization, where it potentially allows the efficient addition of new categories during classification. Results show that classification accuracy still requires improvements, but models generated from one domain are shown to be effectively able to be reused in a different one.