10081 research outputs found
Sort by
Advanced analytical methods for profiling disease-related alterations in diabetic patients with kidney impairment
Get PDFIn pursuing personalized therapeutic approaches, understanding molecular and biomolecular alterations in disease states is crucial for elucidating pathophysiological mechanisms and discovering predictive or diagnostic biomarkers. Investigations on complex samples eventually demand a combination of analytical approaches. The choice of the most appropriate and informative analytical approach depends on sample complexity and on the investigation’ aim/question, often requiring a multi-technique strategy for robust insights.
In this PhD project, high-resolution mass spectrometry (HRMS) and surface plasmon resonance (SPR) were exploited to investigate structural and functional alterations in human serum albumin (HSA) within diseases characterized by enhanced oxidative stress and inflammatory response, with a specific focus on diabetic kidney disease (DKD). HRMS analyses in DKD patients revealed reduced levels of native albumin with concomitant increased levels of oxidized forms (mainly). The levels of both effective (eHA) and reduced HSA (rHA) forms dose-dependently correlated with renal dysfunction, showing higher diagnostic accuracy than total albumin (tHA) in DKD diagnosis. Moreover, eHA emerged as an independent predictor of DKD.
To investigate whether structural alterations impact HSA binding properties, SPR technology was employed. Novel biosensing surfaces functionalized with a selective anti-HSA antibody were developed and validated for immunocapturing HSA directly from patients’ plasma to achieve a sensing surface snapshotting patient-specific HSA heterogeneity. Additionally, the biosensing surface facilitated the evaluation of novel bioresponsive prodrugs for tailored delivery systems targeting the tumor site specifically.
Furthermore, the investigation delved into DKD-related metabolic changes using an untargeted metabolomic approach. The study highlighted progressive alterations in metabolic profiles, shedding light on potential key metabolites and pathways affected in DKD.
In conclusion, the combination of HRMS and SPR technologies provided comprehensive insights into molecular changes associated with DKD. The final aim is to identify additional biomarkers for clinical use and deepen understanding of disease-related metabolic alterations, with potential applications for precise therapeutic interventions
Exploring the psychometric dimensions: theoretical and practical applications of artificial intelligence and gamification in education, learning, and neuropsychological assessment
Get PDFIn recent decades, the intersection of artificial intelligence (AI) and gamification has reshaped learning, education, and neuropsychological assessment. Tracing its roots to mid-20th-century pioneers like Alan Turing, this convergence reflects historical progress. As personal computing emerged in the late 20th century, AI principles integrated into education alongside the rising interest in gamification. Titled "Exploring Psychometric Dimensions: Theoretical and Practical Applications of Artificial Intelligence and Gamification in Education, Learning, and Neuropsychological Assessment," this dissertation aims to delve into the theoretical foundations and real-world applications of incorporating artificial intelligence and gamification in the realms of education, learning, and neuropsychological assessment with a psychometric perspective. The integration of AI and gamification goes beyond traditional pedagogical methods, promising personalized educational experiences. AI's data analysis, machine learning, and adaptive algorithms complement gamification's game design elements, engaging learners and enhancing motivation. In neuropsychological assessment, this fusion offers an innovative framework for evaluating cognitive functions with the potential for more accurate evaluations and enhanced participant engagement. Structured into five main sections, the dissertation begins with a clear introduction, outlining the two primary themes: artificial intelligence and gamification. Part I, "Theoretical Contributions," explores how AI influences human development and essential skills in education and examines gamification's knowledge and constraints. Part II, "Applications on Neuropsychological Assessment and Psychometrics", delves into the integration of AI and gamification in cognitive assessment. Part III, "Applications on Training", focuses on AI's role in training and recommender systems in learning. The dissertation concludes with a general discussion summarizing findings and suggesting future directions. Through critical examination and interdisciplinary exploration, this research contributes to understanding the transformative impact of AI and gamification on learning, education, and neuropsychological assessment practices
Development of machine learning data pipeline and applications to agricultural and livestock structures
Get PDFBig data analysis has made its way up to being one of the most requested skill in today’s world. Industries and research endeavours increasingly rely on advanced analytics techniques to navigate the vast volume of data originating from monitoring systems, wearables, cameras, and similar sources. The sheer amount of data collected has surpassed the feasibility of manual analysis or traditional statistical methods. Moreover, recent years have witnessed a significant reduction in the cost of powerful computing machines, enabling widespread access to advanced models for processing extensive datasets. The confluence of this accessible computational power and the inherent flexibility of machine learning has resulted in a thriving landscape for data analysis. Nowadays, machine learning technique are being applied in every sector, from physics to finance, from agriculture to medicine. In general, data science entails the ability to extract information, and eventually even knowledge, from large amount of data, often without the constraints of a meticulously designed experimental setup. This work delves into the applications of data science, specifically focusing on precision livestock farming, monitoring systems, and enhancing energy efficiency in agro-industrial buildings
Novel techniques for harnessing symbolic and structured information into machine learning
Get PDFIn recent years, we have assisted to a new spring of Artificial Intelligence (AI). This transformation has been characterized by a shift from the symbolic methods prevalent in the last century to a focus on sub-symbolic techniques, driven by the remarkable achievements of deep learning in areas such as computer vision and natural language processing.
Despite the successes of sub-symbolic, data-driven methods, recent years have seen a growing inclination towards hybrid models that synergize symbolic and sub-symbolic approaches. This trend stems from several inherent limitations in purely data-driven systems. Firstly, these systems often redundantly learn concepts that are already part of common knowledge or are well-understood by domain experts. Secondly, data-driven methods may struggle to adhere to specific constraints, such as those dictated by natural laws or user-imposed rules, whereas symbolic methods can manage these constraints more easily. Lastly, the black-box nature of most sub-symbolic methods poses challenges in terms of interpretability and explainability, in contrast to the more transparent symbolic approaches.
In the context of machine learning and deep learning, these challenges have given rise to the emergent field of informed machine learning. This new domain aims to exploit the strengths of both symbolic and sub-symbolic methods by formalizing and incorporating existing task-specific knowledge into traditional machine learning workflows.
The core objective of this thesis is to explore and advance the field of informed machine learning. It presents innovative algorithms within this domain and conducts a thorough investigation of existing methodologies. The applications of these algorithms are explored in two significant areas of AI: predictive modeling and decision support systems. To validate the practical utility of these algorithms, the thesis undertakes a comprehensive empirical evaluation. The findings from these studies provide concrete evidence of the effectiveness of informed machine learning solutions in addressing the highlighted challenges
Brand language in the social media marketplace
Get PDFLanguage is one of the core elements of branding. While marketing scholars have investigated brand language in various traditional marketing contexts such as advertising, little is known about the social media realm. With the present monographic dissertation, the author aims to offer new insights in this regard. This work focuses on internet slang, a form of language often used by both consumers and brands in the social media marketplace. Should brands use internet slang in an attempt to enhance brand relevance, or should they avoid it to reduce consumer inferences of co-optation? Bridging classic theories on impression management with the study of language, this dissertation starts by proposing a new conceptualization of internet slang. Most importantly, it studies the consequences of brands’ adoption of this language using a multi-method investigation that combines text analysis of thousands of field data, statistical modeling, and controlled preregistered experiments. The findings illustrate an inverted U-shaped relationship between internet slang intensity and consumer responses. The curve is flattened by higher brand trust, higher brand coolness, and non-promotional message goal. Furthermore, the studies demonstrate that the curvilinear effect of internet slang intensity on consumer responses is driven by two simultaneous underlying mechanisms: On one hand, increasing internet slang intensity elicits perceptions of message playfulness; on the other hand, high internet slang intensity heightens consumers’ persuasion knowledge. These results offer new insights into brand language and the debate concerning brands capitalizing on or opposing trends. Additionally, this dissertation represents the first work to offer actionable insights regarding the optimal internet slang intensity to be used in marketing messages, thus assisting marketers in crafting successful communications
Computational fluid dynamics applied to indoor and outdoor environments for green and smart production processes.
Get PDFThis thesis explores the application of Computational Fluid Dynamics (CFD) to enhance green and smart production in various environmental conditions, demonstrating its versatility in microclimate management across indoor agricultural and outdoor urban area. This work is important for environmental management, urban planning, and agricultural sciences.
The first study focuses on greenhouse microclimate control, crucial for optimizing crop growth and productivity. It employs a porous media model based on Darcy's law to examine thermal interactions within a greenhouse cultivating sweet peppers, integrating energy balance considerations, mass transfer and heat exchange. This approach significantly advances modeling and assessment of greenhouse climates, enhancing indoor agricultural environments.
The second study investigates the aerodynamic properties of Basil and Mentuccia, using advanced image processing to quantify Leaf Area Density (LAD). These measurements allow for a precise differentiation between the leafy areas and the porous spaces within the crops, which is crucial for an accurate aerodynamic analysis. The insights extend beyond theoretical aerodynamics, offering practical applications in agricultural planning and plant-environment interactions.
The final study applies CFD to urban settings, analyzing how trees and water bodies influence wind patterns and temperature in an urban park in Igualada, Barcelona. It integrates a porous media model with a water–air interface exchange model in three dimensional simulations. The model highlights significant temperature reductions at pedestrian levels, emphasizing the role of integrated green systems in reducing urban heat islands and enhancing urban comfort.
Overall, this thesis underscores CFD's potential as a tool for sustainable development, with broad implications for improving indoor agricultural methods and urban green infrastructure planning, contributing to a more sustainable future
Essays in Bayesian macroeconometrics
Get PDFThe first chapter introduces a "theory-coherent" shrinkage prior for Time-Varying Parameters Vector Autoregressive (TVP-VAR) models. The Theory Coherent TVP-VAR (TC-TVP-VAR) significantly enhances inference precision and forecast accuracy over standard TVP-VAR models. This approach demonstrates superior predictive capabilities for GDP growth and inflation and provides more accurate impulse response analyses, especially in assessing the impacts of macroeconomic shocks during the Zero Lower Bound (ZLB) period.
The second chapter proposes a VAR model with stochastic volatility and time-varying skewness, aimed at monitoring macroeconomic tail risks. It develops an MCMC sampler for Bayesian estimation of VARs with Skew-Normal and Skew-t shocks, featuring stochastic volatility and time-varying skewness. The chapter shows that these models often outperform semi-parametric methods like quantile regression in forecasting macroeconomic risks.
The third chapter applies a Bayesian VAR model with stochastic volatility and time-varying skewness to assess labor risks in the euro area and the United States. It explores the asymmetry of the shocks to the unemployment rate in relation to real activity and financial risk factors. The model is also used to study stagflation risks, emphasizing labor market dynamics in the inflation-unemployment trade-off
Grassroots fights against corruption in the digital age: infrastructural activism in Italy and Spain
Get PDFThis thesis examines the challenges and opportunities of the digital age for anti-corruption activism, casting light on a specific form of activism, called infrastructural activism. This form of activism expands the repertoire of action and contention of civil society actors by creating the preconditions for the development, maintenance, and diffusion of different types of socio-technical infrastructures: grassroots and institutionalized whistleblowing infrastructures, on the one hand, and community and platform-based monitoring infrastructures, on the other.
Indeed, the research considers two types of initiatives, located between Italy and Spain, that over time have incorporated different types of digital technologies into their anti-corruption practices for two main purposes. First, to facilitate whistleblowing by implementing the adoption of open-source software that guarantees high standards of security and anonymity. Second, to monitor governmental actors and combat institutional opacity through the use of public data, open databases, and data-driven platforms.
Bridging corruption and social movement studies with science and technology, media, and journalism studies, this thesis identifies a specific perspective for looking at the grassroots anti-corruption struggle in platform and datafied societies. Indeed, infrastructural activism and related infrastructures play a pivotal role in this struggle by anchoring not only the efforts of civil society organizations but also the efforts of other actors who may use (or replicate) them to detect corruption or prevent it by increasing transparency.
Thus, adopting the lens of infrastructural activism, this dissertation contributes to social movement studies by encouraging a debate on how the broader process of “infrastructuralization of platform-based services”, in which platforms acquire certain characteristics of infrastructures, can also affect platforms and technologies developed by grassroots collective actors
Advancing abstractive long-input summarization in low-resource regimes: methods, datasets and benchmarks
Get PDFThe field of natural language processing (NLP) has experienced remarkable growth in recent years, driven by the emergence of transformer-based models. In today's information-rich era, automatic text summarization has become a pivotal pursuit within NLP. It aims to distill complex textual documents into their essential elements, providing valuable support to experts engaged in labor-intensive tasks. Technically, extractive summarization identifies and extracts key sentences from the source text, whereas abstractive summarization amalgamates, rephrases, and paraphrases essential topics from the input to generate the synthesis. However, significant advances are often impeded by resource constraints. State-of-the-art solutions face challenges in low-resource regimes, relying on substantial computational power and extensive collections of annotated examples for effective training. This heavy dependence on resources poses a significant obstacle for small and medium-sized organizations, limiting their ability to manage these unsustainable costs. This thesis delves into the challenges of text summarization in low-resource environments, presenting multiple techniques categorized by methodology to advance abstractive summarization of long inputs. We show that text segmentation, retrieval-augmented generation, and knowledge injection and distillation are integral components of the solution, whose effectiveness is complemented by the availability of new rigorous datasets and benchmarks. Overall, this work represents a significant step towards real-world practical solutions that do not necessitate reliance on extensive manually curated data and high-memory GPUs.Il campo dell'elaborazione del linguaggio naturale (NLP) ha registrato una crescita notevole negli ultimi anni, supportata dall'introduzione di nuovi modelli basati sul modello transformer. In un'epoca caratterizzata da un incessante flusso di informazioni, il compito di sintesi automatica del testo ha assunto un ruolo di particolare rilevanza all'interno del NLP. Questa attività mira a sviluppare sistemi in grado di generare sintesi concise e significative di documenti testuali, supportando gli esperti in compiti intensivi. Tecnicamente, la sintesi estrattiva identifica ed estrae frasi chiave dal testo originale, mentre la sintesi astrattiva aggrega e riformula i temi essenziali dell'input per generare il riassunto. Tuttavia, il suo successo è spesso oscurato dalla crescente richiesta di risorse computazionali. Le soluzioni all'avanguardia soffrono i regimi a basse risorse, poiché fanno affidamento su una potenza di calcolo sostanziale e su ampie collezioni di esempi annotati per un addestramento efficace. Questa forte dipendenza dalle risorse costituisce un ostacolo significativo per le organizzazioni di piccole e medie dimensioni, limitando la loro capacità di gestire questi costi insostenibili. Questa tesi approfondisce le sfide della sintesi del testo in ambienti a bassa risorsa, presentando diverse tecniche categorizzate per metodologia per avanzare la sintesi astrattiva di lunghi documenti. Mostriamo che la segmentazione del testo, la generazione con recupero e l'iniezione e distillazione della conoscenza sono componenti integrali della soluzione, la cui efficacia è integrata dalla disponibilità di nuovi dataset e benchmark rigorosi. Nel complesso, questo lavoro rappresenta un passo significativo verso soluzioni pratiche nel mondo reale che non richiedono una dipendenza da numerosi dati curati manualmente e da GPU con memoria elevata
Community structure and dynamics of sharks and their relatives in the Adriatic sea
Get PDFChondrichthyans—sharks, rays, and chimaeras—represent one of the most evolutionarily distinct and ecologically diverse groups of vertebrates. However, overfishing poses a severe threat to this group, with one-third of global chondrichthyan species at risk of extinction. The Mediterranean Sea is a critical hotspot for this risk.
In my research, I address the urgent need for comprehensive knowledge about chondrichthyan species distribution and their interaction with fisheries in the Adriatic Sea—a heavily exploited basin within the Mediterranean.
By analyzing long-term fishery-independent monitoring data and employing state-of-the-art distribution modeling techniques, I uncover key insights.
The chondrichthyan community exhibits strong depth and spatial sorting. Some viviparous species are abundant in the shallower northern Adriatic, while there is a west-to-east gradient of increasing abundance and species richness.
Over time, dominant chondrichthyan species have increased in abundance, yet at different rates, except for the threatened spurdog, which continues to decline. Consequently, the community now displays a lower community-weighted age at first maturity and fraction of viviparous species compared to the 1990s.
To estimate the spatial overlap between bottom trawling and chondrichthyans, I introduce a novel approach that combines Vessel Monitoring System data (representing fishing effort) with geostatistical species distribution models.
By employing diverse overlap metrics that capture different levels of organization—from individual species to the entire community—I estimate substantial spatial overlap between intensive trawling, areas of high species richness, and the presence of threatened chondrichthyans (i.e. assessed as Vulnerable, Endangered and Critically Endangered by the IUCN Red List). These areas of high overlap are primarily concentrated in the northern and central offshore sectors of the western Adriatic, highlighting specific areas of conservation concern.
In summary, this thesis contributes innovative statistical frameworks and ecological insights and may help in guiding conservation efforts to sustain chondrichthyan populations in the Adriatic Sea and beyond