An investigation of imputation methods for discrete databases and multi-variate time series

When it comes to information sets in real life, often pieces of the whole set may not be available. This problem can find its origin in various reasons, describing therefore different patterns. In the literature, this problem is known as Missing Data. This issue can be fixed in various ways, from not taking into consideration incomplete observations, to guessing what those values originally were, or just ignoring the fact that some values are missing. The methods used to estimate missing data are called Imputation Methods. The work presented in this thesis has two main goals. The first one is to determine whether any kind of interactions exists between Missing Data, Imputation Methods and Supervised Classification algorithms, when they are applied together. For this first problem we consider a scenario in which the databases used are discrete, understanding discrete as that it is assumed that there is no relation between observations. These datasets underwent processes involving different combina- tions of the three components mentioned. The outcome showed that the missing data pattern strongly influences the outcome produced by a classifier. Also, in some of the cases, the complex imputation techniques investigated in the thesis were able to obtain better results than simple ones. The second goal of this work is to propose a new imputation strategy, but this time we constrain the specifications of the previous problem to a special kind of datasets, the multivariate Time Series. We designed new imputation techniques for this particular domain, and combined them with some of the contrasted strategies tested in the pre- vious chapter of this thesis. The time series also were subjected to processes involving missing data and imputation to finally propose an overall better imputation method. In the final chapter of this work, a real-world example is presented, describing a wa- ter quality prediction problem. The databases that characterized this problem had their own original latent values, which provides a real-world benchmark to test the algorithms developed in this thesis.

Studio del danno indotto dalla detonazione e della temperatura media di pistoni in lega di alluminio e sviluppo di un modello analitico per il calcolo in Real-Time della temperatura dei gas di scarico

In campo motoristico, negli ultimi anni, la ricerca si è orientata allo studio approfondito dell'efficienza di combustione, individuandone in primo luogo i principali aspetti limitanti. Primo tra tutti la detonazione che, essendo dannosa per i componenti del motore (e in particolare quelli della camera di combustione), è adesso al centro di molti studi. L'obiettivo è di conoscerla a fondo in modo da poterne arginare gli effetti. Questa tesi si colloca in un ampio progetto volto a perseguire tale risultato. Infatti, lo studio del danno che viene indotto sui componenti della camera di combustione (i pistoni in particolare), della sua morfologia, della localizzazione prevalente e i principali parametri ai quali esso risulta correlabile, fanno parte dell'attività sperimentale esposta in questo lavoro. Essa si concentra inoltre sul degrado termico della lega dei pistoni a seguito di prove a banco sul motore, che si pongono l'obiettivo di provocare elevati livelli di detonazione e su eventuali benefici che derivano dal poterne accettare episodi di entità incipiente. A tale proposito, viene esposto e validato un modello di temperatura dei gas di scarico Real Time, tramite il quale è possibile calcolare la temperatura di essi una volta noto il punto motore.

Digital Forensics AI: on Practicality, Optimality, and Interpretability of Digital Evidence Mining Techniques

Digital forensics as a field has progressed alongside technological advancements over the years, just as digital devices have gotten more robust and sophisticated. However, criminals and attackers have devised means for exploiting the vulnerabilities or sophistication of these devices to carry out malicious activities in unprecedented ways. Their belief is that electronic crimes can be committed without identities being revealed or trails being established. Several applications of artificial intelligence (AI) have demonstrated interesting and promising solutions to seemingly intractable societal challenges. This thesis aims to advance the concept of applying AI techniques in digital forensic investigation. Our approach involves experimenting with a complex case scenario in which suspects corresponded by e-mail and deleted, suspiciously, certain communications, presumably to conceal evidence. The purpose is to demonstrate the efficacy of Artificial Neural Networks (ANN) in learning and detecting communication patterns over time, and then predicting the possibility of missing communication(s) along with potential topics of discussion. To do this, we developed a novel approach and included other existing models. The accuracy of our results is evaluated, and their performance on previously unseen data is measured. Second, we proposed conceptualizing the term “Digital Forensics AI” (DFAI) to formalize the application of AI in digital forensics. The objective is to highlight the instruments that facilitate the best evidential outcomes and presentation mechanisms that are adaptable to the probabilistic output of AI models. Finally, we enhanced our notion in support of the application of AI in digital forensics by recommending methodologies and approaches for bridging trust gaps through the development of interpretable models that facilitate the admissibility of digital evidence in legal proceedings.

Optimizing AI at the Edge: from network topology design to MCU deployment

The first topic analyzed in the thesis will be Neural Architecture Search (NAS). I will focus on two different tools that I developed, one to optimize the architecture of Temporal Convolutional Networks (TCNs), a convolutional model for time-series processing that has recently emerged, and one to optimize the data precision of tensors inside CNNs. The first NAS proposed explicitly targets the optimization of the most peculiar architectural parameters of TCNs, namely dilation, receptive field, and the number of features in each layer. Note that this is the first NAS that explicitly targets these networks. The second NAS proposed instead focuses on finding the most efficient data format for a target CNN, with the granularity of the layer filter. Note that applying these two NASes in sequence allows an "application designer" to minimize the structure of the neural network employed, minimizing the number of operations or the memory usage of the network. After that, the second topic described is the optimization of neural network deployment on edge devices. Importantly, exploiting edge platforms' scarce resources is critical for NN efficient execution on MCUs. To do so, I will introduce DORY (Deployment Oriented to memoRY) -- an automatic tool to deploy CNNs on low-cost MCUs. DORY, in different steps, can manage different levels of memory inside the MCU automatically, offload the computation workload (i.e., the different layers of a neural network) to dedicated hardware accelerators, and automatically generates ANSI C code that orchestrates off- and on-chip transfers with the computation phases. On top of this, I will introduce two optimized computation libraries that DORY can exploit to deploy TCNs and Transformers on edge efficiently. I conclude the thesis with two different applications on bio-signal analysis, i.e., heart rate tracking and sEMG-based gesture recognition.

Relazione tra carica virale, reazione immunitaria dell'ospite, microbiota e decorso clinico della bronchiolite acuta nei bambini con età inferiore ai 12 mesi (RMB2018)

Il principale motivo di ospedalizzazione del lattante è la bronchiolite acuta, causata in primis dal Virus Respiratorio Sinciziale (VRS). In questo ambito appaiono controversi i risultati in letteratura riguardo: carica del VRS, risposta immunitaria ed infiammatoria dell’ospite, microbiota del paziente (durante l’infezione e per successivo sviluppo di wheezing e asma). In questo studio di coorte prospettico monocentrico vengono arruolati lattanti peraltro sani ricoverati per primo episodio di bronchiolite acuta, da VRS o da altro agente, per valutare primariamente la relazione tra decorso clinico e carica del VRS, secondariamente l’associazione con specifiche composizioni e modifiche nel tempo del microbiota nasofaringeo ed intestinale durante la fase acuta e nel lungo termine in relazione a sviluppo di wheezing ricorrente. Nello studio sono stati arruolati finora 89 pazienti, di cui 68 con bronchiolite da VRS (76.4%), con analisi della carica virale su 41 lattanti (60.3%) e del microbiota su 20 (29.4%). L’analisi della carica del VRS non ha riscontrato associazione tra outcome di severità clinica quali necessità e durata di ossigenoterapia, nonché durata di ricovero. La presenza di trend di associazione tra carica virale all’ingresso e picco del VRS-RNA con necessità di ossigenoterapia ad alto flusso (HFNC) e l’associazione significativa di clearance di VRS con HFNC (p = 0.03) suggeriscono che la carica virale potrebbe influenzare la severità della bronchiolite da VRS. Le analisi del microbiota evidenziano numerosi genera mai descritti finora in letteratura a nostra conoscenza (es. Alloiococcus e Leptotrichia su aspirato, nonché tutti i dati emersi dallo studio della saliva, prima volta in letteratura) ed un generale aumento della diversity a 6 mesi dalla dimissione. Occorrerà completare l’analisi della carica del VRS con quella del sistema infiammatorio dell’ospite, nonché studiare il microbiota sui campioni del follow-up a lungo termine per la verifica di eventuali associazioni con sviluppo di wheezing e asma.

Real-time anomaly detection of gamma-ray bursts for the Cherenkov Telescope Array using deep learning

The Cherenkov Telescope Array (CTA) will be the next-generation ground-based observatory to study the universe in the very-high-energy domain. The observatory will rely on a Science Alert Generation (SAG) system to analyze the real-time data from the telescopes and generate science alerts. The SAG system will play a crucial role in the search and follow-up of transients from external alerts, enabling multi-wavelength and multi-messenger collaborations. It will maximize the potential for the detection of the rarest phenomena, such as gamma-ray bursts (GRBs), which are the science case for this study. This study presents an anomaly detection method based on deep learning for detecting gamma-ray burst events in real-time. The performance of the proposed method is evaluated and compared against the Li&Ma standard technique in two use cases of serendipitous discoveries and follow-up observations, using short exposure times. The method shows promising results in detecting GRBs and is flexible enough to allow real-time search for transient events on multiple time scales. The method does not assume background nor source models and doe not require a minimum number of photon counts to perform analysis, making it well-suited for real-time analysis. Future improvements involve further tests, relaxing some of the assumptions made in this study as well as post-trials correction of the detection significance. Moreover, the ability to detect other transient classes in different scenarios must be investigated for completeness. The system can be integrated within the SAG system of CTA and deployed on the onsite computing clusters. This would provide valuable insights into the method's performance in a real-world setting and be another valuable tool for discovering new transient events in real-time. Overall, this study makes a significant contribution to the field of astrophysics by demonstrating the effectiveness of deep learning-based anomaly detection techniques for real-time source detection in gamma-ray astronomy.

On the Democratization of AI: Bridging the Gap with AutoML

Riding the wave of recent groundbreaking achievements, artificial intelligence (AI) is currently the buzzword on everybody’s lips and, allowing algorithms to learn from historical data, Machine Learning (ML) emerged as its pinnacle. The multitude of algorithms, each with unique strengths and weaknesses, highlights the absence of a universal solution and poses a challenging optimization problem. In response, automated machine learning (AutoML) navigates vast search spaces within minimal time constraints. By lowering entry barriers, AutoML emerged as promising the democratization of AI, yet facing some challenges. In data-centric AI, the discipline of systematically engineering data used to build an AI system, the challenge of configuring data pipelines is rather simple. We devise a methodology for building effective data pre-processing pipelines in supervised learning as well as a data-centric AutoML solution for unsupervised learning. In human-centric AI, many current AutoML tools were not built around the user but rather around algorithmic ideas, raising ethical and social bias concerns. We contribute by deploying AutoML tools aiming at complementing, instead of replacing, human intelligence. In particular, we provide solutions for single-objective and multi-objective optimization and showcase the challenges and potential of novel interfaces featuring large language models. Finally, there are application areas that rely on numerical simulators, often related to earth observations, they tend to be particularly high-impact and address important challenges such as climate change and crop life cycles. We commit to coupling these physical simulators with (Auto)ML solutions towards a physics-aware AI. Specifically, in precision farming, we design a smart irrigation platform that: allows real-time monitoring of soil moisture, predicts future moisture values, and estimates water demand to schedule the irrigation.

Sviluppo di un ambiente virtualizzato per la simulazione di attacchi a reti time-sensitive

Questo lavoro di tesi ha visto come obiettivo finale quello di realizzare una se- rie di attacchi, alcuni di questi totalmente originali, ai protocolli della famiglia Time-Sensitive Networking (TSN) attraverso lo sviluppo di un’infrastruttura virtualizzata. L’infrastruttura è stata costruita e progettata utilizzando mac- chine virtuali con Quick EMUlator (QEMU) come strato di virtualizzazione ed accelerate attraverso Kernel-based Virtual Machine (KVM). Il progetto è stato concepito come Infrastrucutre as Code (IaC), attraverso l’ausilio di Ansible e alcuni script shell utilizzati come collante per le varie parti del progetto.

Generazione efficiente di molteplici scenari di compromissione cyber ai fini di test e formazione

Questo lavoro si propone di implementare tre scenari di compromissione informatica tramite l'ausilio della strumentazione fornita da Ansible e Docker. Dopo una prima parte teorica di presentazione delle più recenti vulnerabilità/compromissioni informatiche, si passa all'illustrazione degli strumenti e dell'architettura degli scenari, anche tramite l'ausilio di codice. Tramite la funzione UNIX time si effettua l'analisi di diverse tecniche di distribuzione, dimostrando come l'automazione abbia effettivi e seri vantaggi rispetto ad una continua implementazione manuale, principalmente da un punto di vista temporale e computazionale.

Tecnologie e workflow per la digitalizzazione e la visualizzazione immersiva e real time del prodotto tessile

Il settore del mobile è frammentato, caratteristica che non ha mai permesso una completa digitalizzazione dei suoi comparti: la comunicazione è risultata spesso difficoltosa, responsabile di incomprensioni e spesso mancanza di puntualità, con conseguente dispendio di risorse. Nel mio progetto di tesi analizzo principalmente la filiera degli imbottiti e quella tessile, collaterale ad essa. Il target di mio interesse sono le aziende produttrici di arredi che si servono di configuratori per la presentazione dei prodotti ai propri clienti. Ho ipotizzato un workflow suddivisibile in due fasi, una attuabile nel presente ed una seconda che implica una variazione attuabile nel futuro prossimo. Per la prima fase ho studiato un workflow agile al fine di digitalizzare le texture a partire da acquisizioni fotografiche. Analizzando il comportamento di vari tessuti ed elaborando un metodo di riproduzione digitale, ho sperimentato una metodologia per una resa superficiale degli imbottiti accurata e fotorealistica sia su piattaforme web che in realtà aumentata. Lo studio nasce dall’esigenza di apportare migliorie agli attuali configuratori e renderli effettivamente utili al fine per cui sono stati pensati: garantire una semplificazione del processo di acquisto rispondendo alle necessità degli utenti di vedere in anteprima l’aspetto finale del proprio arredo, collocato nello spazio abitativo. Il configuratore che ne deriva, restituirà una visualizzazione fedele dei tessuti anche in realtà aumentata. L’intelligenza artificiale fungerà da supporto nel processo decisionale dell’utente e la scelta dei materiali che compongono l’arredo sarà guidata in maniera tale da generare nell’utente un senso di consapevolezza circa gli impatti sull’ambiente. Per la seconda fase propongo una modalità di creazione dei tessuti completamente digitalizzata, a favore di una filiera sostenibile che riesce in questo modo ad abbattere costi e sprechi, con conseguente risparmio di risorse.

Comparison Of Postoperative Stability Of Three Rigid Internal Fixation Techniques After Sagittal Split Ramus Osteotomy For Mandibular Advancement.

The aim of this investigation was to compare the skeletal stability of three different rigid fixation methods after mandibular advancement. Fifty-five class II malocclusion patients treated with the use of bilateral sagittal split ramus osteotomy and mandibular advancement were selected for this retrospective study. Group 1 (n = 17) had miniplates with monocortical screws, Group 2 (n = 16) had bicortical screws and Group 3 (n = 22) had the osteotomy fixed by means of the hybrid technique. Cephalograms were taken preoperatively, 1 week within the postoperative care period, and 6 months after the orthognathic surgery. Linear and angular changes of the cephalometric landmarks of the chin region were measured at each period, and the changes at each cephalometric landmark were determined for the time gaps. Postoperative changes in the mandibular shape were analyzed to determine the stability of fixation methods. There was minimum difference in the relapse of the mandibular advancement among the three groups. Statistical analysis showed no significant difference in postoperative stability. However, a positive correlation between the amount of advancement and the amount of postoperative relapse was demonstrated by the linear multiple regression test (p < 0.05). It can be concluded that all techniques can be used to obtain stable postoperative results in mandibular advancement after 6 months.

Identification Of Corynebacterium Spp. Isolated From Bovine Intramammary Infections By Matrix-assisted Laser Desorption Ionization-time Of Flight Mass Spectrometry.

Corynebacterium species (spp.) are among the most frequently isolated pathogens associated with subclinical mastitis in dairy cows. However, simple, fast, and reliable methods for the identification of species of the genus Corynebacterium are not currently available. This study aimed to evaluate the usefulness of matrix-assisted laser desorption ionization/mass spectrometry (MALDI-TOF MS) for identifying Corynebacterium spp. isolated from the mammary glands of dairy cows. Corynebacterium spp. were isolated from milk samples via microbiological culture (n=180) and were analyzed by MALDI-TOF MS and 16S rRNA gene sequencing. Using MALDI-TOF MS methodology, 161 Corynebacterium spp. isolates (89.4%) were correctly identified at the species level, whereas 12 isolates (6.7%) were identified at the genus level. Most isolates that were identified at the species level with 16 S rRNA gene sequencing were identified as Corynebacterium bovis (n=156; 86.7%) were also identified as C. bovis with MALDI-TOF MS. Five Corynebacterium spp. isolates (2.8%) were not correctly identified at the species level with MALDI-TOF MS and 2 isolates (1.1%) were considered unidentified because despite having MALDI-TOF MS scores >2, only the genus level was correctly identified. Therefore, MALDI-TOF MS could serve as an alternative method for species-level diagnoses of bovine intramammary infections caused by Corynebacterium spp.

Molecular Variations In Aromatic Cosolutes: Critical Role In The Rheology Of Cationic Wormlike Micelles.

Wormlike micelles formed by the addition to cetyltrimethylammonium bromide (CTAB) of a range of aromatic cosolutes with small molecular variations in their structure were systematically studied. Phenol and derivatives of benzoate and cinnamate were used, and the resulting mixtures were studied by oscillatory, steady-shear rheology, and the microstructure was probed by small-angle neutron scattering. The lengthening of the micelles and their entanglement result in remarkable viscoelastic properties, making rheology a useful tool to assess the effect of structural variations of the cosolutes on wormlike micelle formation. For a fixed concentration of CTAB and cosolute (200 mmol L(-1)), the relaxation time decreases in the following order: phenol > cinnamate> o-hydroxycinnamate > salicylate > o-methoxycinnamate > benzoate > o-methoxybenzoate. The variations in viscoelastic response are rationalized by using Mulliken population analysis to map out the electronic density of the cosolutes and quantify the barrier to rotation of specific groups on the aromatics. We find that the ability of the group attached to the aromatic ring to rotate is crucial in determining the packing of the cosolute at the micellar interface and thus critically impacts the micellar growth and, in turn, the rheological response. These results enable us for the first time to propose design rules for the self-assembly of the surfactants and cosolutes resulting in the formation of wormlike micelles with the cationic surfactant CTAB.

Photoelastic Analysis Of All-on-four Concept Using Different Implants Angulations For Maxilla.

Conventional tilted implants are used in oral rehabilitation for heavily absorbed maxilla to avoid bone grafts; however, few research studies evaluate the biomechanical behavior when different angulations of the implants are used. The aim of this study was evaluate, trough photoelastic method, two different angulations and length of the cantilever in fixed implant-supported maxillary complete dentures. Two groups were evaluated: G15 (distal tilted implants 15°) and G35 (distal tilted implants 35°) n = 6. For each model, 2 distal tilted implants (3.5 x 15 mm long cylindrical cone) and 2 parallel tilted implants in the anterior region (3.5 x 10 mm) were installed. Photoelastic models were submitted to three vertical load tests: in the end of cantilever, in the last pillar and in the all pillars at the same time. We obtained the shear stress by Fringes software and found values for total, cervical and apical stress. The quantitative analysis was performed using the Student tests and Mann-Whitney test; p ≥ 0.05. There is no difference between G15 and G35 for total stress regardless of load type. Analyzing the apical region, G35 reduced strain values considering the distal loads (in the cantilever p = 0.03 and in the last pillar p = 0.02), without increasing the stress level in the cervical region. Considering the load in all pillars, G35 showed higher stress concentration in the cervical region (p = 0.04). For distal loads, G15 showed increase of tension in the apical region, while for load in all pillars, G35 inclination increases stress values in the cervical region.

Effect Of Composite Containing An Iodonium Salt On The Bond Strength Of Brackets To Bovine Enamel.

This study investigated the effect of the incorporation of an iodonium salt in experimental composites, on the bond strength of metallic brackets bonded to bovine teeth. Two hundred and seventy bovine teeth were embedded in self-curing acrylic resin and divided into 18 groups (n=15), according to the experimental composite with an iodonium salt at molar concentrations 0 (control), 0.5, or 1%; the light-activation times (8, 20 and 40 s); and the storage times (10 min or 24 h). Metallic brackets were fixed on the tooth surface using experimental composites. Photoactivation was performed with a quartz-tungsten-halogen light-curing unit curing unit for 8, 20 and 40 s. The specimens were stored in distilled water at 37 °C for 10 min or 24 h and submitted to bond strength test at 0.5 mm/min. The data were subjected to three-way ANOVA and Tukey's test (α=0.05). The Adhesive Remnant Index (ARI) was used to classify the failure modes. The shear bond strengths (MPa) at 10 min for light-activation times of 8, 20 and 40 s were: G1 - 4.6, 6.9 and 7.1; G2 - 8.1, 9.2 and 9.9; G3 - 9.1, 10.4 and 10.7; and at 24 h were: G1 - 10.9, 11.1 and 11.7; G2 - 11.8, 12.7 and 14.2; G3 - 12.1, 14.4 and 15.8. There was a predominance of ARI score 3 for groups with 10 min storage time, and ARI score 2 for groups with 24 h storage time. In conclusion, the addition of iodonium salt (C05 and C1) to the experimental composite may increase the bond strength of brackets to bovine enamel using reduced light exposure times.