10081 research outputs found
Sort by
Exploiting and generalizing \\ Epistemic uncertainty in reinforcement learning and planning
Get PDFSolving sequential decision-making problems with complex and non-linear dynamics has been a goal of Artificial Intelligence since the conception of the field. Reinforcement Learning (RL) offers an general framework for solving such problems. Its approach learning by direct interaction with the environment, allowing for speculation on the value of candidate solutions, testing, and counter-factual reasoning, has allowed researchers to achieve remarkable achievements in a multitude of challenging problems both simulated and real-world. Nonetheless, the successful application of RL to new problems requires a large degree of task-specific tuning.
One of the main open challenges in RL remains the exploration-exploitation dilemma. An agent that optimizes a cumulative objective in an unknown environment while learning faces the question of whether to trust the current information gathered and exploit it by executing the best-known strategies or take explorative strategies to gather more information with the hope of finding better strategies. The exploration problem has been thoroughly studied in the literature, and a multitude of solutions have been given for tabular domains or continuous domains with known structure. However, when moving to complex domains where neural networks are employed as function approximators, deep and directed exploration is still a challenge.
In this dissertation, we tackle the exploration problem in RL by proposing Wasserstein TD-Learning (WTD), a novel framework that models the uncertainty over the value function in a model-free manner and propagates it across the state-action space by employing variational updates that allow us enough control over the updates to show some desirable theoretical properties in the tabular setting while allowing the method to be easily scalable in the DeepRL setting. This allows us to adapt WTD in a multitude of different settings by adapting algorithms from the literature to handle the distributional nature of our value function, allowing for deep and directed exploration
Functional analysis of interaction between primary HIV-1 nef alleles and host proteins
Get PDFThe HIV-1 accessory protein Nef plays a pivotal role in viral pathogenesis due to its multifunctional character. Nef facilitates immune evasion by promoting the down-modulation of cell surface molecules, including major histocompatibility complex class I (MHC-I), CD4 and a member of the serine incorporator family, SERINC5. Furthermore, the human peroxisomal thioesterase 8 (ACOT8) has been shown to be an important cellular partner of Nef. However, these mechanisms has not yet been fully understood. This study investigated the impact of primary nef alleles on the modulation of MHC-I, CD4 and SERINC5 and their capability to interact with ACOT8.
26 nef alleles were isolated from ART-naïve HIV-1-positive individuals, corresponding to subtypes B, F1, G, C, and CRF02_AG within the M group, and then cloned into a pIRES-eGFP vector for functional analysis. Receptor modulation of MHC-I and CD4 at an endogenous level was evaluated in HEK293T and/or Jurkat T cells using flow cytometry. Additionally, Nef's capability to modulate SERINC5 surface expression was examined by transfecting HEK293T cells to express nef alleles together with SERINC5. The interaction between Nef and ACOT8 was assessed through a co-immunoprecipitation approach in an ACOT8-mScarlet overexpression context. Lastly, Nef polymorphisms potentially associated with these activities were identified and correlated with their respective functions.
This study showed that the majority of primary nef alleles led to a significant downregulation of MHC-I, CD4, and SERINC5 from the cell surface, that correlated with a high conservation of the known functional domains. Moreover, it demonstrated for the first time an interaction between primary nef alleles and ACOT8, revealing that this interaction was highly variable among viral subtypes and might depend on amino acid residues within the central core domain of Nef. Overall, the obtained data contributed to enhancing the understanding of Nef's biological functions and identified new mechanisms potentially involved in HIV pathogenesis
Evaluation of snoRNA levels in HCC: a study explaining the role played by SNORA74D in HCC development
No full textAmong non-coding genes small nucleolar RNA are recently getting attention in the field of prognosis and diagnosis of cancer including the Hepatocellular Carcinoma. The biological roles of snoRNA in HCC are, but no limited to, involvement in initiation, proliferation, metastasis, apoptosis, and cell cycle. Here in this study we are focusing on getting insight into the relationship between snoRNA and HCC. To this end we received sequenced tissue samples from patients with NASH and HCC. After performing initial bioinformatics analysis we found potential set of candidate snoRNAs i.e. SNORA74D, SNORA80D, and SNORA5C which were differentially expressed. We further verified these potential targets with qPCR analysis and found only SNORA74D to be significantly upregulated in HCC samples in comparison to NASH. In order to determine the mechanistic role we depleted HepG2 cell line from SNORA74D and found its potential role in cell survival and proliferation. TGIRT sequencing was also performed on SNORA74D depleted HepG2 samples to determine the prospective downstream effectors and regulators. Multiple protein coding genes and noncoding RNA are significantly dysregulated by knockdown of SNORA74D. Of note, we also characterized and discovered the presence of different isoforms of SNORA74D both in tissue samples and cell lines. The presence of different type of isoforms in cancer tissues compared with normal tissues further suggest potential involvement of SNORA74D in cancer development and progression
Numerical modelling of flow-like landslides and the related phenomena adopting the improved Smoothed Particle Hydrodynamics (SPH) models
Get PDFFlow-like landslides are a common and serious geologic hazard worldwide that can lead to substantial environmental degradation. They can severely damage infrastructure, including roads, bridges, buildings, and utility lines. Additionally, they may alter the natural landscape, cause deforestation, and contribute to soil erosion. Understanding the dynamic characteristics of this type of landslide is crucial for devising effective mitigation strategies and promoting resilience in vulnerable areas.Numerical modeling plays a crucial role in understanding, predicting, and managing flow-like landslides. It provides valuable insights into the complex dynamics of catastrophic landslide events, allowing researchers, engineers, and policymakers to assess risks, and design mitigation measures. In this PhD thesis, three challenging topics (slope failure, landslide-structure interaction (LSI) and entrainment) are thoroughly analyzed by means of numerical modelling. To investigate the first topic, I first implemented the elastic-plastic constitutive model in the DualSPHysics framework to simulate the mechanical behavior of strain history-dependent materials such as cohesive soils. Furthermore, the failure and deformation processes of a cohesive soil slope without strain softening and that of a sensitive clay slope with strain softening, are studied to validate the proposed SPH model. The multiple failure process of the Caijiapo landslide (China) was analyzed by the developed SPH model. A novel SPH model was proposed to address the second and third topics. Due to the importance of friction in the dynamic simulation of flow-like landslides, an HBP rheology model coupled with the friction dissipation was embedded into the DualSPHysics code. Regarding LSI issues, column collapse and sand flume benchmarks were carried out to verify the performance and capability of the developed SPH model. Then, the propagation of a real debris-flow event occurring in northern Italy and its interaction with check dams installed along the flow path were analyzed
Interference management techniques for non-terrestrial networks in 5G/6G systems"
Get PDFNon-Terrestrial Network can be defined as communication systems that use air-/space-borne platforms to establish communication links among various points on Earth’s surface. In the context of (5G), NTN aims at the use of NTN platforms to enhance the flexibility, adaptability and the resilience of 5G Terrestrial Network system as well as to extend its coverage and provide connectivity to rural and underserved/un-served areas. In 6G, NTN is expected to play an even more pivotal role, potentially forming a significant part of the global communication infrastructure. The unification of TN and NTN in 6G is envisioned to enable global coverage, low-latency communication, and support for massive Internet of Things (IoT) deployments. This unification will require advances
in satellite and aerial communication technologies, as well as seamless integration with terrestrial networks to ensure ubiquitous and reliable connectivity. Furthermore, the use of NTN in 6G is expected to bring about new challenges in terms of network architecture and spectrum management. The scarcity of spectrum resources to meet the demanding requirements of future Beyond 5G (B5G)/6G communication systems is considered one of the main challenges. To this aim, the current state of the art considers several strategies aimed at efficient utilization of the available spectrum
bandwidth through either dynamic spectrum utilization schemes, such as Cognitive Radio solutions, or by fully exploiting the spectrum by decreasing the frequency reuse factor down to Full Frequency Reuse. However, the latter solution introduces
substantial Co-Channel Interference or also known as Inter-Beam Interference, thus calling for the use of advanced interference management techniques at the transmitter side, i.e, at the satellite. This issue is addressed throughout the thesis by introducing interference management schemes and solutions by evaluation MIMO beamforming algorithms
Retrotransposons in human evolution and brain diseases: investigating modern and ancient variability through the lens of mobile elements
Get PDFTransposable Elements (TEs) are mobile genetic elements with the ability to replicate themselves and insert their copies in a new genomic location; notably, sequences derived from TEs make up to 53-60% of the human genome. While the vast majority of TEs are not transpositionally active, they can still play functional roles such as modulating the expression or alternative splicing of nearby genes. Moreover, they can generate potentially inheritable insertions, thereby creating human-specific polymorphisms, that in the last years have been used to study human evolution, variability and susceptibility to medical disorders. Here we present a large-scale in silico study on polymorphic TEs, used as genetic variability markers in modern human populations and ancient genomes. Furthermore, we investigated the possible role of TEs in conferring a risk of developing brain disorders, such as addiction, schizophrenia and Alzheimer’s disease. Our results point toward an important role of polymorphic TEs in: 1) shaping human genome evolution, especially the neural genome and human-specific cognitive abilities; 2) influencing phenotypic variability and adaptation; 3) being the target of recent instances of positive selection or more ancient event of adaptive introgression; 4) conferring risk of developing neuro-psychiatric disorders
Development of flux pump-based supply systems for superconducting coils
Get PDFConventional power supplies for superconducting magnets rely on dissipative power electronic converters and current leads, which are a thermal path connecting room temperature to cryogenic superconductors, as well as an additional source of joule dissipation, thereby imposing a significant heat load on the cooling system.
Flux pumps offer a contactless, low-voltage, and high-current alternative to power electronics exciters and current leads solutions. However, as flux pumps physical mechanism is complex and counterintuitive, their engineering process and development of viable and executive designs have proved to present major challenges.
This Ph.D. project focused on numerical modeling, investigation, and engineering analysis of flux pumps. The research aimed to address gaps in scientific research surrounding the technology and its potential disruptive impact on enabling contactless and efficient energization of superconducting magnets. First, a numerical model was developed and validated against experimental results to unveil the intricacies related to voltage rectification in traveling field flux pumps. The model embeds artificial intelligence methods to address the relationship between critical current and n-value of HTS tapes concerning temperature and magnetic field. A new complete equivalent circuit was derived and exploited for fast system-oriented analysis of flux pumps. The numerical model was further exploited to develop an application-oriented optimal design procedure for traveling field flux pumps, which was tested to design and characterize a HTS dynamo and a linear flux pump for real applications. The study found that the performance of the flux pumps is highly dependent on the rated current and its total resistance and discussed the influence of these two load parameters on the viability of flux pumps.
Overall, this Ph.D. project aimed to overcome the empirical approach with which flux pumps have been designed and analyzed for decades and applied a systematic engineering method to facilitate its implementation in real applications
Digital storytelling and open air museum in an urban regeneration project (UIA04-137 DARE): the memory and cultural heritage of the Darsena of Ravenna.
Get PDFGli open air museum sono una tipologia museale nata nel XIX secolo come risposta all'idea più tradizionale di museo. Fin dai loro primi esempi, i musei all’aperto si sono posizionati come contesti in cui veniva valorizzato il patrimonio materiale e immateriale strettamente legato alle comunità di riferimento e alla sua memoria culturale. Nella presente ricerca viene approfondita la nascita degli open air museum e altre sperimentazioni a cielo aperto (ecomusei, archaeological open air museum, musei diffusi, urban open air museum), analizzandone le radici condivise e ciò che li distingue. Viene analizzato il rapporto tra memoria culturale-patrimonio culturale e narrazione. Necessario a identificare i processi di valorizzazione che portano la comunità a un attivo coinvolgimento nella sua conservazione. Queste sono state le premesse per la creazione di strumenti di digital storytelling per la valorizzazione del patrimonio culturale del quartiere Darsena di Ravenna. Il quartiere è stato oggetto del progetto europeo DARE-UIA nato per promuovere nuove metodologie di governance digitale e per stimolare azioni di rigenerazione urbana. Il progetto DARE ha dato l'occasione per riflettere sul rapporto tra forme museali legate a contesti industriali urbani e nuove modalità digitali per valorizzare la memoria culturale e il patrimonio immateriale del quartiere. La ricerca teorica ha portato alla progettazione dell’open air museum della Darsena di Ravenna. L’obiettivo è stato creare una infrastruttura museale che promuova processi di rigenerazione urbana. La metodologia ha previsto la realizzazione di strumenti di digital storytelling, utilizzando tecnologie GIS e 3D per la valorizzazione di un percorso museale diffuso. L’obiettivo è creare una metodologia sostenibile, replicabile in contesti simili.Open air museums are a type of museum that was born in the 19th century as a response to the more traditional idea of a museum. Since their first examples, open-air museums have positioned themselves as contexts where the material and immaterial heritage is closely linked to the communities of reference, and their cultural memory is enhanced. In this research, the birth of open air museums and other open-air experiments (ecomuseums, archaeological open air museums, widespread museums, and urban open air museums) is explored in depth, analysing their shared roots and what distinguishes them. The relationship between cultural memory-cultural heritage and narration is analysed. It is necessary to identify the enhancement processes that lead the community to active involvement in its conservation. These were the premises for creating digital storytelling tools to enhance the cultural heritage of the Darsena district of Ravenna. The neighbourhood was the subject of the European DARE-UIA project created to promote new digital governance methodologies and to stimulate urban regeneration actions. The DARE project allowed to reflect on the relationship between museum forms linked to urban industrial contexts and new digital ways to enhance the neighbourhood's cultural memory and intangible heritage. The theoretical research led to the design of the open air museum of the Darsena in Ravenna. The objective was to create a museum infrastructure that promotes urban regeneration processes. The methodology involved the creation of digital storytelling tools, using GIS and 3D technologies to enhance a widespread museum itinerary. The objective is to create a sustainable methodology replicable in similar contexts
A framework to quantify the 3d left atrium wall motion model on a patient specific basis, in atrial fibrillation patients
Get PDFAtrial fibrillation (AF) is the most common, clinically significant cardiac arrhythmia. AF causes structural remodeling to the left atrium (LA) chamber and reduces the contractility of the walls of the LA and disrupts the normal blood flow within the LA chamber and consequently gives rise to thrombogenic events. Therefore, the primary goal of this study is to propose a step-by-step workflow on a patient-specific basis, which can potentially assess and quantify the global and regional impact of AF on the contractility of LA and left atrial appendage (LAA) as 90% of the clot formation happens within the LAA.
To conduct this study, the analysis was performed separately for the LA and LAA. However, for both the analysis, same cohort of computed tomography (CT) dynamic imaging dataset was utilized. In the case of the LA, a standardized regionalization approach was proposed to define distinct LA regions and then in the next step, global and regional contraction parameters for the LA were defined and computed to characterize LA function. Moreover, for the LAA, a semi-automatic LAA centerline extraction technique was developed and by considering the centerline as reference, LAA global and regional parameters were assessed.
Results showed the potential of the proposed methods to assess and quantify the dysfunctional global and regional LA and LAA mechanical contraction in AF patients. By computing the contraction parameters, global contraction impairment, significant dilation and regional heterogeneity in contraction were noticed in AF patients as compared to the healthy subjects. Additionally, irregular and asynchronous contraction between different LA regions was also noticed in AF patients.
In conclusion, the computed patient-specific contraction indices of LA and LAA were found to have the capabilities to identify and quantify global and regional contraction dysfunctionalities in AF patients
Noninvasive evaluation of the Fontan associated liver disease
Get PDFIntroduzione:
L’intervento di Fontan comporta un aumento istantaneo della pressione venosa centrale che causa, nel medio-lungo termine, una forma di epatopatia specifica detta FALD. Il monitoraggio della FALD è complesso ma potrebbe consentire di bloccarne o rallentarne l’insorgenza. Lo studio ha valutato l’efficacia delle modalità di monitoraggio non invasivo.
Materiale e metodi:
Sei pazienti (età media 24 anni) operati presso l’IRCCS Azienda Ospedaliero Universitaria di Bologna sono stati sottoposti a RMN 4D-Flow e ad Ecodoppler epatico. Sono stati raccolti i dati anagrafici, morfologici, anamnestici e i markers sierologici per il calcolo degli scores MELD-XI, APRI, FIB4, i valori di Shear Stress assiale e circonferenziale e gli indici di pulsatilità e resistenza delle arterie epatica e renale.
Risultati:
Il tempo trascorso tra la Fontan e lo studio è stato di 17,8 anni. Età media alla Fontan 6,8 anni. Tutti i pazienti avevano un quadro compatibile con epatopatia. I markers sierologici e gli scores MELD-XI,APRI e FIB4 si sono dimostrati di scarsa utilità. All’ecografia tutti i pazienti avevano ecostruttura irregolare, splenomegalia e valori elevati di pulsatilità e resistenza dell’arteria epatica e splenica. La rigidità epatica media è stata di 12,4 Kpa. Alla RMN 4DF lo Shear stress assiale è stato massimo a livello del condotto (0,16 Pa) e minimo a livello delle vene sovra epatiche (0,05 Pa). Lo Shear Stress si è mostrato massimo nei pazienti con emodinamica sfavorevole e peggior quadro ecografico addominale, evidenziando aree di inefficienza energetica.
Conclusioni:
La combinazione delle diagnostiche di imaging non invasive potrebbe rivelarsi adeguata per il monitoraggio della FALD. In particolare, la RMN 4D Flow potrebbe rivelare aree di inefficienza energetica predisponenti alla FALD. Questo potrebbe indirizzare in modo specifico la terapia dei pazienti operati o addirittura indurre la modifica del disegno della Fontan verso forme più efficienti.Introduction:
Fontan surgery results in an instantaneous increase in central venous pressure that causes, in the medium to long term, a specific form of hepatopathy called FALD. Monitoring of FALD is complex but could allow its onset to be stopped or slowed. This study evaluated the effectiveness of noninvasive monitoring modalities.
Material and methods:
Six patients (mean age 24 years) operated on at IRCCS Azienda Ospedaliero Universitaria di Bologna underwent 4D-Flow MRI and hepatic Ecodoppler. Biographical, morphological, anamnestic data and serological markers were collected to calculate MELD-XI, APRI, FIB4 scores, axial and circumferential Shear Stress values, hepatic structure and dimension, portal and caval vein dimensions and flow pattern, and hepatic and renal arterial pulsatility and resistance indices.
Results:
The mean age at Fontan was 6.8 years, and the time elapsed between Fontan and the study was 17.8 years. All patients had a picture compatible with hepatopathy. Serologic markers and MELD-XI, APRI, and FIB4 scores proved to be of little use. On ultrasonography, 100% of patients had irregular ecostructure and splenomegaly; one had a nodule <1cm. The mean liver stiffness was 12.4 Kpa. Hepatic and splenic pulsatility and resistance index were elevated in 100% of patients. On MRI, the discrepancy between venous flow and flow values in the pulmonary arteries was 6.1%. Axial shear stress was highest at the level of the duct (0.16 Pa) and lowest at the level of the suprahepatic veins (0.05 Pa). Shear stress seemed to correlate with echo and laboratory findings.
Conclusions:
The combination of noninvasive imaging diagnostics might prove sufficient for monitoring FALD. In particular, 4D Flow MRI could reveal areas of energy inefficiency predisposing to FALD. This could specifically target the therapy of operated patients or even induce the modification of Fontan design toward more efficient forms