1731 research outputs found
Sort by
Efficient and Effective Solutions for Video Classification
Get PDFThe aim of this PhD thesis is to make a step forward towards teaching computers to understand videos in a similar way as humans do. In this work we tackle the video classification and/or action recognition tasks. This thesis was completed in a period of transition, the research community moving from traditional approaches (such as hand-crafted descriptor extraction) to deep learning. Therefore, this thesis captures this transition period, however, unlike image classification, where the state-of-the-art results are dominated by deep learning approaches, for video classification the deep learning approaches are not so dominant. As a matter of fact, most of the current state-of-the-art results in video classification are based on a hybrid approach where the hand-crafted descriptors are combined with deep features to obtain the best performance. This is due to several factors, such as the fact that video is a more complex data as compared to an image, therefore, more difficult to model and also that the video datasets are not large enough to train deep models with effective results. The pipeline for video classification can be broken down into three main steps: feature extraction, encoding and classification. While for the classification part, the existing techniques are more mature, for feature extraction and encoding there is still a significant room for improvement. In addition to these main steps, the framework contains some pre/post processing techniques, such as feature dimensionality reduction, feature decorrelation (for instance using Principal Component Analysis - PCA) and normalization, which can influence considerably the performance of the pipeline. One of the bottlenecks of the video classification pipeline is represented by the feature extraction step, where most of the approaches are extremely computationally demanding, what makes them not suitable for real-time applications. In this thesis, we tackle this issue, propose different speed-ups to improve the computational cost and introduce a new descriptor that can capture motion information from a video without the need of computing optical flow (which is very expensive to compute). Another important component for video classification is represented by the feature encoding step, which builds the final video representation that serves as input to a classifier. During the PhD, we proposed several improvements over the standard approaches for feature encoding. We also propose a new feature encoding approach for deep feature encoding. To summarize, the main contributions of this thesis are as follows3: (1) We propose several speed-ups for descriptor extraction, providing a version for the standard video descriptors that can run in real-time. We also investigate the trade-off between accuracy and computational efficiency;
(2) We provide a new descriptor for extracting information from a video, which is very efficient to compute, being able to extract motion information without the need of extracting the optical flow; (3) We investigate different improvements over the standard encoding approaches for boosting the performance of the video classification pipeline.;(4) We propose a new feature encoding approach specifically designed for encoding local deep features, providing a more robust video representation
The importance of climatic and ecological factors for vector-borne infections: Culex pipiens and West Nile virus
No full textAbout three quarters of human emerging infectious diseases are caused by zoonotic pathogens, and many of them are spread by vectors such as mosquitoes. Mathematical models nowadays represent very powerful tools to make investigations and predictions for biological dynamical systems, providing helpful insights that can be extremely valuable for several aims. In this thesis, we will focus on a particular mosquito-borne zoonosis, West Nile virus (WNV), a flavivirus of emerging public health relevance in Europe and North America, and its main European vector, Culex pipiens mosquitoes. As the transmission of mosquito-borne diseases is largely driven by the abundance of the vector, to design appropriate control strategies it is crucial to understand the population dynamics of existing vector populations and evaluate how it depends on biotic and environmental factors. This thesis presents some new mathematical models that provide insights on several aspects of mosquito population dynamics by using different statistical and computational approaches, including for instance Linear Models and Markov chain Monte Carlo technique. Specifically, they aim to study the effect of biotic and abiotic factors on Cx. pipiens dynamics by using adult mosquito trapping data, gathered over several years in Northern Italy, to feed theoretical models. Furthermore, the effects of host competition and vector feeding preferences on the dynamics of a vector-borne infection (such as WNV) are investigated through a more theoretical study
Mechanical Modelling of single and collective cells behavior
No full textRecent experimental results have suggested important direct implications of viscoelasticity of human cells and cell cytoskeleton dynamics on some relevant collective and at single-cell behaviors such as migration, adhesion, and morphogenesis. Other experimental studies have been performed on individual cancer and healthy cells of different types, demonstrating that the former were about 70% softer than the latter. In this thesis with the aim of characterizing — and gaining insights into — the frequency response of single-cell systems to mechanical stimuli (typically LITUS), a generalized viscoelastic paradigm which combines classical and spring-pot based (fractional derivative) models is presented. Than the modelling has been enriched considering the non-linear effect of the prestress, induced in protein filaments during cell adhesion and in the cell membrane (with a simple multiscale scheme that incorporates finite elasticity and a 3-D circus tent-like model), on the overall cell stiffness and finally determining its influence on the in-frequency response of the cell. The theoretical results have shown that the differences in stiffness — at least in principle — allow us to mechanically discriminate between tumor and normal cells: the critical frequencies associated with oscillation magnitude peaks (from tens to hundreds of kilohertz) could be helpfully utilized for targeting or ad hoc altering the functions of cancer cells. An experimental validation of the theoretical results is an ongoing work and the preparation of the experimental setup is also presented.
In this thesis some first models have been presented to replicate in-vivo collective behavior of cells. Coherent angular rotation of epithelial cells has been reproduced by a cell-centered based mechanical model in which units are polarized, motile, and interact with the neighboring cells via harmonic forces. Starting from this model a continuum non-linear viscoelastic model incorporating the dynamics of liquid crystals has been studied and some preliminary numerical simulations have been performed
Collective oscillations of a trapped atomic gas in low dimensions and thermodynamics of one-dimensional Bose gas
No full textUltracold atoms are exceptional tools to explore the physics of quantum matter. In fact, the high degree of tunability of ultracold Bose and Fermi gases makes them ideal systems for quantum simulation and for investigating macroscopic manifestations of quantum effects, such as superfluidity.
In ultracold gas research, a central role is played by collective oscillations. They can be used to study different dynamical regimes, such as superfluid, collisional, or collisionless limits or to test the equation of state of the system. In this thesis, we present a unified description of collective oscillations in low dimensions covering both Bose and Fermi statistics, different trap geometries and zero as well as finite temperature, based on the formalism of hydrodynamics and sum rules. We discuss the different behaviour exhibited by the second excited breathing mode in
the collisional regime at low temperature and in the collisionless limit at high temperature in a 1D trapped Bose gas with repulsive contact interaction. We show how this mode exhibits a single-valued excitation spectrum in the collisional regime and two different frequencies in the collisionless limit. Our predictions could be important for future research related to the thermalization and damping phenomena in this low-dimensional system. We show that 1D uniform Bose gases exhibit a non-monotonic temperature dependence of the chemical potential characterized by an increasing-with-temperature behaviour at low temperature. This is due to the thermal excitation of phonons and reveals an interesting analogy with the behaviour of superfluids.
Finally, we investigate a gas with a finite number N
of atoms in a ring geometry at T = 0. We discuss explicitly the deviations of the thermodynamic behaviour in the ring
from the one in the large N limit
Managing the Scarcity of Monitoring Data through Machine Learning in Healthcare Domain
No full textIn the field of Ubiquitous Computing, a significant problem of building accurate machine learning models is the effort and time consuming process to gather labeled data for the learning algorithm. Moreover, efficient data use demands are constantly growing. These demands for efficient data use are growing constantly. Researchers are therefore exploring the use of machine learning techniques to overcome the problem of data scarcity. In healthcare, classification tasks require a ground truth normally provided by an expert physician, ending up with a small set of labeled data with a larger set of unlabeled data. It is also common to rely on self-reported data through questionnaires, however, this introduce an extra burden to the user who is not always able or willing to fill in. Finally, in some healthcare domains it is important to be able to provide immediate response (feedback), even if the user is not familiarized with the use of an application. In all of these cases the amount of available data may be insufficient to produce reliable models. This thesis proposes a new approach specifically designed for the challenges in producing better predictive models. We propose using our novel Intermediate Models to predict the mood variables associated with the questionnaire using data acquired from smartphones. Then, we use the predicted mood variables with the rest of the data to predict the class, in our empirical assessment, the state mood of a bipolar disorder patient or stress levels of employees have been used. The motivation behind this new approach is that there are relevant proposed methods such as latent variables used as intermediate information helping to create better predictive models. These methods are used in literature to complete the missing data using the most common value, the most probable value given the class, or induce a model for predicting missing values using all the information from features and the class. However, these variables are artificially created and used as intermediate information to build better model. In our Intermediate Models, we know in advance how many mood variables to use and we have the information from these variables, which allow us to produce better models. To address scarce data, we propose applying a semi-supervised learning setting while taking advantage of the presence of all unlabeled datasets. In addition, we propose using transfer learning methods that is used to improve the learning performance with the aim at avoiding expensive data labeling efforts. To the best of our knowledge, there are few works that have used transfer learning for healthcare applications to address the problem of limited labeled data. The proposed methods have been applied in two different healthcare fields: mental-health and human behaviour field. This thesis addresses two classification problems, a) classification of episodic state of bipolar disorder patients, and b) detecting work-related stress using data acquired from smartphone sensing modalities
Security assessment of open source third-parties applications
Get PDFFree and Open Source Software (FOSS) components are ubiquitous in both proprietary and open source applications. In this dissertation we discuss challenges that large software vendors face when they must integrate and maintain FOSS components into their software supply chain. Each time a vulnerability is disclosed in a FOSS component, a software vendor must decide whether to update the component, patch the application itself, or just do nothing as the vulnerability is not applicable to the deployed version that may be old enough to be not vulnerable. This is particularly challenging for enterprise software vendors that consume thousands of FOSS components, and offer more than a decade of support and security fixes for applications that include these components.
First, we design a framework for performing security vulnerability experimentations. In particular, for testing known exploits for publicly disclosed vulnerabilities against different versions and software configurations.
Second, we provide an automatic screening test for quickly identifying the versions of FOSS components likely affected by newly disclosed vulnerabilities: a novel method that scans across the entire repository of a FOSS component in a matter of minutes. We show that our screening test scales to large open source projects.
Finally, for facilitating the global security maintenance of a large portfolio of FOSS components, we discuss various characteristics of FOSS components and their potential impact on the security maintenance effort, and empirically identify the key drivers
Machine Learning for Investigating Post-Transcriptional Regulation of Gene Expression
No full textRNA binding proteins (RBPs) and non-coding RNAs (ncRNAs) are key actors in post-transcriptional gene regulation. By being able to bind messenger RNA (mRNA) they modulate many regulatory processes. In the last years, the increasing interest in this level of regulation favored the development of many NGS-based experimental techniques to detect RNA-protein interactions, and the consequent release of a considerable amount of interaction data on a growing number of eukaryotic RBPs. Despite the continuous advances in the experimental procedures, these techniques are still far from fully uncovering, on their own, the global RNA-protein interaction system. For instance, the available interaction data still covers a small fraction (less than 10%) of the known human RBPs. Moreover, experimentally determined interactions are often noisy and cell-line dependent. Importantly, obtaining genome-wide experimental evidence of combinatorial interactions of RBPs is still an experimental challenge. Machine learning approaches are able to learn from the data and generalize the information contained in them. This might give useful insights to help the investigation of the post-transcriptional regulation. In this work, three machine learning contributions are proposed. They aim at addressing the three above-mentioned shortcomings of the experimental techniques, to help researchers unveiling some yet uncharacterized aspects of post-transcriptional gene regulation.
The first contribution is RNAcommender, a tool capable of suggesting RNA targets to unexplored RBPs at a genome-wide level. RNAcommender is a recommender system that propagates the available interaction data, considering biologically relevant aspects of the RNA-protein interactions, such as protein domains and RNA predicted secondary structure.
The second contribution is ProtScan, a tool that models RNA-protein interactions at a single-nucleotide resolution. Learning models from experimentally determined interactions allows to denoise the data and to make predictions of the RBP binding preferences in conditions that are different from those of the experiment.
The third and last contribution is PTRcombiner, a tool that unveils the combinatorial aspects of post-transcriptional gene regulation. It extracts clusters of mRNA co-regulators from the interaction annotations, and it automatically provides a biological analysis that might supply a functional characterization of the set of mRNAs targeted by a cluster of co-regulators, as well as of the binding dynamics of different RBPs belonging to the same cluster
Computational models for impact mechanics and related protective materials and structures
Get PDFThe mechanics of impacts is not yet well understood due to the complexity of materials behaviour under extreme stress and strain conditions and is thus of challenge for fundamental research, as well as relevant in several areas of applied sciences and engineering. The involved complex contact and strain-rate dependent phenomena include geometrical and materials non-linearities, such as wave and fracture propagation, plasticity, buckling, and friction. The theoretical description of such non-linearities has reached a level of advance maturity only singularly, but when coupled -due to the severe mathematical complexity- remains limited. Moreover, related experimental tests are difficult and expensive, and usually not able to quantify and discriminate between the phenomena involved. In this scenario, computational simulation emerges as a fundamental and complementary tool for the investigation of such otherwise intractable problems. The aim of this PhD research was the development and use of computational models to investigate the behaviour of materials and structures undergoing simultaneously extreme contact stresses and strain-rates, and at different size and time scales. We focused on basic concepts not yet understood, studying both engineering and bio-inspired solutions. In particular, the developed models were applied to the analysis and optimization of macroscopic composite and of 2D-materials-based multilayer armours, to the buckling-governed behaviour of aerographite tetrapods and of the related networks, and to the crushing behaviour under compression of modified honeycomb structures. As validation of the used approaches, numerical-experimental-analytical comparisons are also proposed for each case
The G3P Framework: guiding the design process of games for purpose
No full textVideo games were created more than half a century ago and have rapidly become part of the common culture. Hundreds of video games are developed each year, differentiated by elements such as genre (e.g., puzzle or action), technology (e.g., virtual reality or handheld consoles) and audience (e.g., casual or hardcore gamers). Over the years, interest has grown in using video games for more serious purposes, intended to teach, train or inform. These games are designed exploiting a playful component to stimulate users to achieve specific goals. In general, video games have a complex design consisting of various interrelated elements and strongly influenced by the needs of players. The addition of a serious purpose increases this complexity, making the design of an effective product more challenging. This thesis focuses on the development of video games for purpose, proposing the G3P, a framework to drive the process of design. The G3P instantiates different game design theories in practice, adapting them according to a participatory approach in order to structure the design process. In this renewed perspective, different stakeholders are involved, contributing to the definition and articulation of the various elements of the game. In favour of this process the author presents the development of Skies of Manawak, a video game for neurocognitive training targeted at children with dyslexia. The design and implementation of this video game ran over two years and was divided into several stages. From the first conceptualisation to the release of the game, the stakeholders - domain experts and players - were involved. Each phase was aimed at the extension of the design and the revision of the overall work. The results collected from workshops and testing activities allowed assessing the effectiveness of Skies of Manawak in terms of player experience and cognitive training. In this regard, the analysis of the related work grounded the G3P framework; the various design activities held during the development of Skies of Manawak completed it. This manuscript represents a contribution to game design theories of video games for purpose, supporting the theoretical foundation of this research topic, still young but rapidly growing
Poesia senza verso. La poesia in prosa in Italia
Get PDFQuesto lavoro ripercorre la storia della poesia in prosa in Italia nel Novecento. . Inizialmente vengono considerate le interazioni fra prosa e poesia in seguito all’influenza della traduzione di opere straniere nell’Ottocento, parallelamente alla nascita del poemetto in prosa e del verso libero. Quindi viene fornito un inquadramento storico della koiné letteraria di inizio Novecento presso la quale la poesia in prosa diventa una forma privilegiata. La poesia in prosa viene considerata il genere della modernità italiana: ciò è spiegato attraverso la ricostruzione del contesto letterario (la polemica contro il romanzo, la nascita del verso libro, l’ibridazione dei generi letterari). Infine, la poesia in prosa viene esaminata alla luce delle teorie sul modernismo italiano. Le opere esaminate in modo più analitico appartengono a cinque autori, considerabili case study: Giovanni Boine, Dino Campana, Pietro Jahier, Camillo Sbarbaro, Scipio Slataper.L’ultima parte della tesi è dedicata alla poesia in prosa a partire dagli anni Settanta. Il quinto capitolo è di tipo monografico, ed è incentrato sull’opera di Giampiero Neri. Gli ultimi due capitoli hanno di nuovo un taglio storico: il decennio che va dal 2001 al 2010 è considerato più in dettaglio, specularmente a quanto già fatto per il periodo 1908-1919