10081 research outputs found
Sort by
The role of the ventromedial prefrontal cortex in self- and event-related schemata
Get PDFThe ventromedial prefrontal cortex (vmPFC) is one of the largest regions of the human brain. Its activity has been linked to a plethora of cognitive functions, such as self-referential cognition, mind wandering, memory recollection, and many more. Yet, we still lack an exhaustive understanding of the essence of its functional specialization, if any. The schema theory proposes that vmPFC’s overarching role lies in the activation of schematic knowledge in neocortex, which is then used by other brain regions, according to environmental demands. Therefore, the present thesis addresses the question of whether a vmPFC damage degrades schema-mediated cognition. In Chapter 1 I begin to investigate the role of vmPFC in imparting the memory advantage for self-referential information in memory. Whilst healthy and brain-damaged controls exhibit superior recall for self- (vs other-) related items, such advantage is proven absent in vmPFC patients, suggesting a degradation of the self-schema. In Chapter 2 I aim to clarify whether this lack of self-referential prioritization stems from vmPFC patients’ memory deficits, or from an impairment in self-knowledge itself. Whilst healthy and brain-damaged controls exhibit more consistent self- rather than other-referential ratings over time, vmPFC patients do not, again suggesting a disturbance of the self-schema. In Chapter 3 we investigate vmPFC’s involvement in activating (reinstatement) and using (instantiation) event-schemata, demonstrating that a vmPFC damage hinders schema reinstatement, wherein vmPFC patients reinstate incomplete, nebulous schemas. Finally, in Chapter 4 we model frontal and posterior cortical interactions in a hybrid Potts model of cortical dynamics, revealing a capacity of the frontal cortex to act as the source of predominant influence on latching dynamics. We interpret the result considering the schematic influence exerted from the frontal lobe on posterior brain regions observed in experimental practice. Finally, by modelling a frontal lesion, we reproduce event construction deficits of vmPFC damaged patients
Geometry projection for additively manufacture variable stiffness continuous fiber-reinforced polymer structures—A unified topology optimization approach for multi-layered composite laminates
Get PDFContinuous fiber fused filament fabrication (CF4) is a layer-by-layer technique used to print carbon fiber-reinforced polymers (CFRPs) with a spatial in-plane variation of the fiber orientation, thus offering great flexibility in fabricating variable-stiffness CFRP laminates (VS-CFRP-Ls). However, not only is the design of VS-CFRP-Ls unintuitive, but the material directionality also introduces a nonconvex design space further amplified by the various VS-CFRP-Ls' design parameters. Designing multi-layered VS-CFRP-Ls, therefore, requires advanced computational design tools---such as topology optimization based on the geometry projection method---to take full advantage of the design freedom compatible with CF4.
This thesis addresses these challenges by developing computational tools for optimizing multilayered VS-CFRP-Ls. Unlike constant stiffness composites, VS-CFRP-Ls lack analytical formulations, necessitating discretization techniques like finite element analysis. The research develops and investigates several topology optimization formulations to streamline the design process, considering CF4's manufacturing constraints and material distribution strategies. The method reduces design variables by employing geometry projection within TO while ensuring manufacturability. Extensions of this approach cater to additive manufacturing requirements, yielding multilayered VS-CFRP-L designs with enhanced mechanical properties. Numerical examples demonstrate the efficacy of the proposed methodology in achieving stiffness-driven VS-CFRP-Ls designs, which can be manufactured using conventional and additive manufacturing processes
the neural based of predictive styles along the autism-schizophrenia continuum
Get PDFThe present thesis aimed to push the boundaries of understanding on the pivotal role played by neural oscillations in modulating Bayesian inference and decision-making processes. Additionally, it explored how inter-individual differences in Autistic and Schizotypal traits shape the neuro-behavioral mechanisms underpinning perceptual inference. A primary contribution of this work is the development of a novel bio-behavioral model of the Autism Spectrum Disorder (ASD) - Schizophrenia Spectrum Disorder (SSD) continuum. This model conceptualizes both conditions as disorders of predictive abilities, stemming from disruptions in brain oscillatory patterns. Building on foundational insights, the thesis employs advanced computational models and state-of-art EEG methodologies to elucidate the distinct oscillatory signatures of predictive processing and understand the rhythmic underpinnings of maladaptive predictive inference in ASD and SSD. Specifically, the research investigated 1) the brain connectivity patterns related to ASD-SSD continuum, 2) the behavioral and electrophysiological mechanisms of prior knowledge integration in decision-making, and 3) how dispositional factors associated with ASD and SSD traits direct individual predictive strategies in both laboratory settings and real-world scenarios. These studies provided a comprehensive understanding of cognitive styles and brain oscillatory codes governing perception, decision-making, and health-related attitudes, contributing significantly to our knowledge of how these mechanisms underlie manifestations observable in the ASD-SSD continuum
Novel language models and methods for semantic representation learning, self-supervised retrieval, and summarization
Get PDFUnderstanding natural language represents one of the most fascinating and complex challenges that has engaged hundreds of researchers over the last century. The richness and variety of language, used daily in a myriad of contexts by people of every age and profession, make its processing by computers a goal of significant relevance. From writing an email to reading a newspaper, listening to a podcast, or simply interacting on social media, language emerges as the universal medium of communication, making the development of systems capable of understanding and naturally interacting with users crucial.
The creation of interfaces that speak and understand human language represents not only a step forward in the field of human-machine interaction but also has the potential to revolutionize our way of living, making technology more accessible and intuitive for everyone, breaking down barriers, and reducing learning curves.
Today, we witness significant progress in this field, with the development of advanced chatbots, large language models, and diffusion models that are transforming the way we interact with technology. It is already possible, for example, to generate detailed images from a simple description, design entire websites, or receive personalized advice by expressing one's needs in their preferred language.
However, at the heart of this transformation lies a fundamental question, which is the subject of this thesis: how can we teach computers to truly understand the meaning hidden behind human words?
In this thesis work, we delve into the analysis of this complex human-machine relationship, examining the current limitations of neural networks, such as transformers, and addressing concrete obstacles, such as the lack of annotated data. Moreover, we introduce new strategies aimed at mitigating, if not resolving, these challenges.Comprendere il linguaggio naturale rappresenta una delle sfide più affascinanti e complesse che ha coinvolto centinaia di ricercatori nell'ultimo secolo. La ricchezza e la varietà del linguaggio, utilizzato quotidianamente in una miriade di contesti da persone di ogni età e professione, rendono la sua elaborazione da parte dei computer un obiettivo di grande rilevanza. Dalla scrittura di un'email alla lettura di un giornale, dall'ascolto di un podcast alla semplice interazione sui social media, il linguaggio emerge come il mezzo comunicativo universale, rendendo cruciale lo sviluppo di sistemi in grado di comprenderlo e di interagire naturalmente con l'utente.
La realizzazione di interfacce che parlano e comprendono la lingua umana non solo rappresenta un passo avanti nel campo dell'interazione uomo-macchina, ma ha il potenziale di rivoluzionare il nostro modo di vivere, rendendo la tecnologia più accessibile e intuitiva per tutti, abbattendo barriere e riducendo le curve di apprendimento.
Oggigiorno, assistiamo a significativi progressi in questo ambito, con lo sviluppo di chatbot avanzati, large language models e diffusion models che trasformano il modo in cui interagiamo con la tecnologia. È già possibile, ad esempio, generare immagini dettagliate partendo da una semplice descrizione, progettare interi siti web o ricevere consigli personalizzati esprimendo le proprie esigenze nella lingua di preferenza.
Tuttavia, al cuore di questa trasformazione giace una questione fondamentale, che costituisce l'oggetto di questa tesi: come possiamo insegnare ai computer a comprendere realmente il significato nascosto dietro le parole umane?
In questo lavoro di tesi, ci addentriamo nell'analisi di questa complessa relazione uomo-macchina, esaminando i limiti attuali delle reti neurali, come i transformers, e affrontando ostacoli concreti, quali la carenza di dati annotati. Inoltre, introdurremo nuove strategie volte a mitigare, se non risolvere, queste problematiche
Analysis of Eurocentrism in fashion. Between appropriation, stereotypes, and new design approaches.
Get PDFQuesto studio si propone di investigare varie tematiche riguardanti la moda, al fine di approfondire la comprensione dei meccanismi sottostanti all'appropriazione culturale e ad altre forme di insensibilità presenti nel settore, con l'obiettivo di esplorare le complessità culturali e sociali della moda contemporanea attraverso una mappatura e un'analisi delle tendenze creative.
Utilizzando dati provenienti dall'analisi delle sfilate e strumenti di intelligenza artificiale per il riconoscimento delle immagini, si intende evidenziare le forme di insensibilità culturale presenti nella moda tra il 2010 e il 2020, promuovendo un dibattito critico sulla decolonizzazione e sull'inclusione nel settore della moda. Attraverso una revisione critica della letteratura e l'analisi dei dati raccolti, questa ricerca mira a contribuire alla comprensione delle dinamiche di appropriazione culturale e a promuovere un approccio più etico e sensibile alla creazione e rappresentazione nella moda.
In particolare, si propone di offrire un quadro completo delle pratiche creative ripetute adottate dai marchi di moda e dei significati socioculturali da essi generati, evidenziando le culture maggiormente esposte ad atteggiamenti eurocentrici. Nonostante l'intero sistema moda sia impegnato in un graduale processo di trasformazione positiva, questo lavoro si propone di sollecitare una riflessione critica e un'azione orientata verso una moda più inclusiva e rispettosa della diversità culturale.This study aims to investigate various issues regarding fashion, in order to deepen the understanding of the underlying mechanisms of cultural appropriation and other forms of insensitivity present in the industry, with the objective of exploring the cultural and social complexities of contemporary fashion through mapping and analysis of creative trends. By utilizing data from runway analysis and artificial intelligence tools for image recognition, it aims to highlight forms of cultural insensitivity in fashion between 2010 and 2020, fostering critical debate on decolonization and inclusion in the fashion sector. Through a critical literature review and analysis of collected data, this research aims to contribute to the understanding of cultural appropriation dynamics and to promote a more ethical and sensitive approach to creation and representation in fashion. Specifically, it seeks to provide a comprehensive framework of repeated creative practices adopted by fashion brands and the socio-cultural meanings they generate, highlighting cultures most exposed to Eurocentric attitudes. Despite the entire fashion system being engaged in a gradual process of positive transformation, this work aims to stimulate critical reflection and action oriented towards a more inclusive and culturally respectful fashion
Embedding AI into constrained devices: a multi-faceted approach
Get PDFThanks to their data-driven nature, machine and deep learning approaches have recently reached super-human performance, promoting the last artificial intelligence (AI) spring. Accordingly, the application of such techniques to the industrial world has vastly grown in popularity. However, the most common deep learning models, namely neural networks (NNs), are characterised by an intrinsic trade-off between performance and efficiency. Focusing on raw performance, recent efforts produced highly complex NN models made of several millions or even billions of parameters. This complexity hinders the application of AI into industrial devices and appliances characterised by limited computational capabilities and resources. Accordingly, in this thesis, we focus on the embedding AI into constrained devices problem, to which we refer to as the open research challenge of applying AI techniques to devices characterised by limited computational capabilities and resources. We tackle the embedding AI task reframing the problem from a NN efficientisation perspective, where the aim is the minimisation of the resource usage of NNs, either during their optimization process or their deployment phase. We propose a pioneer multi-faceted approach in which we consider both (i) the available efficientisation approaches – aiming at analysing and overcoming some of their limitations –, and (ii) to leverage Neuro-Symbolic integration (NeSy) mechanisms to tackle the efficientisation perspective. As a result of our twofold perspective, we shed new light on the NN efficientisation issue, highlighting the groundbreaking opportunities available leveraging NeSy systems
The settlement of the middle valley of Aso during the early stages of romanisation, evolution and role of sanctuaries.
Get PDFIl progetto di ricerca mira ad approfondire le indagini del gruppo di ricerca dell’Università di Bologna nella regione medio-adriatica. Lo studio ha coinvolto due livelli di indagine: uno locale, incentrato sul territorio di Monte Rinaldo, e uno territoriale, che abbraccia le province di Ascoli Piceno e Fermo, attraverso la catalogazione delle evidenze archeologiche note da fonti bibliografiche e archivistiche.
L'obiettivo principale è stato esplorare le modalità di insediamento nel territorio circostante al santuario di Monte Rinaldo, con un'enfasi particolare sulle prime fasi della romanizzazione e la loro evoluzione nei primi secoli dell’età romana.
Nell'area della media valle dell’Aso, attorno al santuario di “La Cuma”, sono state effettuate ricognizioni di superficie per verificare le informazioni provenienti dai legacy-data. Successivamente, l'attenzione si è spostata su un'area poco esplorata, con l'obiettivo di rivelare il popolamento di epoca romana legato all'area sacra e le sue trasformazioni nel tempo. Una volta completata la fase di acquisizione, i dati raccolti sono stati analizzati sia dal punto di vista della cultura materiale sia in un contesto spaziale, utilizzando informazioni geografiche e geomorfologiche.
L'interazione tra i vari dati, gestiti attraverso una piattaforma GIS, ha permesso di identificare aree insediative preferenziali in diversi contesti cronologici. Il progetto ha inoltre beneficiato della continuazione delle indagini stratigrafiche iniziate nel santuario repubblicano di Monte Rinaldo, che, salvo ulteriori specifiche verifiche, si sono concluse nel 2023.The research project aims to deepen the investigations carried by the research group of the University of Bologna. This study mainly involved two levels of investigation: one local, focused on the territory of Monte Rinaldo, and one territorial, encompassing the provinces of Ascoli Piceno and Fermo, through a first cataloguing of archaeological evidence known from bibliographic and archival sources. The main objective was to explore the methods of settlement in the territory surrounding the sanctuary of Monte Rinaldo, with particular emphasis on the early stages of Romanization and their evolution in the early centuries of the Roman Age. In the area of the middle valley of the Aso, near the sanctuary of "La Cuma", archaeological surveys were carried out to verify information from legacy-data. Subsequently, the focus shifted to an area little explored, with the aim of revealing the settlement of Roman times linked to the sacred area and its transformations over time. Once the acquisition phase was completed, the collected data were analysed both from the point of view of material culture and in a spatial context, using geographical and geomorphological information. The interaction between the various data, managed through a GIS platform, has allowed to identify preferential settlement areas in different chronological contexts. The project also benefited from the stratigraphic investigations in the Republican sanctuary of Monte Rinaldo, which, subject to further specific checks, ended with the 2023 campaign
Essays in public economics: behavioural responses to taxes, public good provision, and social welfare
Get PDFThis thesis presents empirical and theoretical contributions to three key topics in the field of public economics: (i) individual behavioural responses to taxation, (ii) optimal public good provision and taxation, and (iii) social welfare theory. The first chapter of the thesis analyses the responses of self-employed individuals to the incentives of the tax system in Italy. I exploit the discontinuity created by the eligibility threshold of the preferential turnover tax regime to estimate how self-employed individuals adjust turnover -- i.e. revenues -- in response to taxes. By combining bunching techniques and a newly developed theoretical framework describing the individual choice between a turnover tax regime and profit-based tax system, I estimate the elasticity of turnover in three sectors of the economy: professional services, retail and accommodation and business intermediaries. The second chapter presents a theory of optimal provision of a (risky) public good when individuals have heterogeneous preferences for risk. People with different attitudes to risk have different views on the extent to which society should invest in certain risky projects. I investigate how these different views should be taken into account for the choice of the optimal policy. The choice of the welfare criterion to use to evaluate the optimal policy is a related, but separate, issue that is explored in the third chapter. Governments are often required to make decisions under risk. However, it is not clear how society should evaluate such choices when individuals have different attitudes to risk. The third chapter proposes a new welfare criterion to evaluate social options in the presence of risk and heterogeneous attitudes to it, and it applies to the specific case of constant relative risk aversion utility function
Exploring the effect of RAD51/BRCA2 inhibition to pursue synthetic lethality with PARPi in in-vitro models of pancreatic cancer
Get PDFSynthetic lethality (SL) is an innovative framework for discovering novel anticancer treatments for personalized targeted therapies. Two genes are synthetically lethal if the inhibition of either gene alone has no effect on cell viability, but their simultaneous impairment leads to cell death. In this context, the Food and Drug Administration approved in 2014 the PARP inhibitor (PARPi) olaparib for oncology patients with BRCA1/2 mutations. My Ph.D. research project is focused on pancreatic cancer, an oncological need. It is aimed at exploiting a new paradigm, dubbed “fully small-molecule-induced synthetic lethality”, which was already presented in previous studies by the research group of Professors A. Cavalli (Italian Institute of Technology) and M. Roberti (University of Bologna). It is based on the possibility of triggering SL by using only small molecules: a PARPi and a RAD51/BRCA2 disruptor that mimics the BRCA2-defective condition. RAD51 and BRCA2 are two key proteins in the homologous recombination (HR) pathway. Their interaction is mediated by eight motifs of BRCA2, among which the fourth (BRC4) has the highest affinity for RAD51. The first section of this thesis describes the effects on cell cultures of a synthetic BRC4 peptide, which reproduced the expected outcomes of HR pathway inhibition, such as the reduction of RAD51 nuclear foci following DNA damage and the increased response to chemotherapeutic agents. A subsequent proteomic study in BRC4-exposed cultures led to identify a statistically significant downregulation of three proteins (FANCI, FANCD2, RPA3) involved in the DNA damage response. The second section of the thesis is focused on the biological characterization of compound 46, a RAD51/BRCA2 inhibitor identified by IIT colleagues. 46 was studied on both 2D and 3D pancreatic cancer models in combination with the PARPi talazoparib. Taken together, the obtained results suggested that the talazoparib-46 combination is a potential inducer of SL
Language policies in a multilingual context: a comparative analysis between Italy and China
Get PDFQuesta ricerca ha l’obiettivo di confrontare le politiche linguistiche in Italia e in Cina, due Paesi multilingue, con particolare attenzione alla legislazione riguardante le lingue comuni. Nella prima parte della tesi viene descritta e confrontata la situazione sociolinguista dei due Paesi, dal punto di vista sia diacronico sia sincronico, permettendo così di tracciare un quadro generale della panoramica sociolinguistica, la quale costituisce il fondamento per analizzare le politiche linguistiche adottate. La seconda parte della tesi descrive e analizza rispettivamente la legislazione linguistica sulla lingua comune dei due Paesi, mostrando la diversità e le caratteristiche delle pratiche legislative di Cina e Italia su questo tema. L’ultima parte della tesi presenta un confronto tra le legislazioni linguistiche sulle lingue comuni da diverse prospettive, individuando e discutendo differenze e punti in comune. La tesi si conclude con un’analisi delle motivazioni che spingono la Cina e l’Italia ad adottare scelte differenti, mettendo in luce la diversità delle politiche linguistiche e la complessità dei fattori che le determinano.This research aims to compare language policies in Italy and China, two multilingual countries, with a particular focus on legislation concerning the common language. The first part of the thesis describes and compares the sociolinguistic situation in the two countries, from both a synchronic and a diachronic perspective, thus allowing for a general outline of the sociolinguistic overview, which is the foundation for analyzing the adopted language policies. The second part of the thesis describes and analyzes the language legislation concerning the common language in both countries, highlighting the diversity and characteristics of China's and Italy's legislative practices. The last part of the thesis presents a comparison of language legislations on common languages from various perspectives, identifying and discussing differences and commonalities. The thesis concludes with an analysis of the motivations driving China and Italy to make different choices, highlighting the diversity of language policies and the complexity of the factors that determine them