Towards Unstructured Knowledge Integration in Natural Language Processing

In the last decades, Artificial Intelligence has witnessed multiple breakthroughs in deep learning. In particular, purely data-driven approaches have opened to a wide variety of successful applications due to the large availability of data. Nonetheless, the integration of prior knowledge is still required to compensate for specific issues like lack of generalization from limited data, fairness, robustness, and biases. In this thesis, we analyze the methodology of integrating knowledge into deep learning models in the field of Natural Language Processing (NLP). We start by remarking on the importance of knowledge integration. We highlight the possible shortcomings of these approaches and investigate the implications of integrating unstructured textual knowledge. We introduce Unstructured Knowledge Integration (UKI) as the process of integrating unstructured knowledge into machine learning models. We discuss UKI in the field of NLP, where knowledge is represented in a natural language format. We identify UKI as a complex process comprised of multiple sub-processes, different knowledge types, and knowledge integration properties to guarantee. We remark on the challenges of integrating unstructured textual knowledge and bridge connections with well-known research areas in NLP. We provide a unified vision of structured knowledge extraction (KE) and UKI by identifying KE as a sub-process of UKI. We investigate some challenging scenarios where structured knowledge is not a feasible prior assumption and formulate each task from the point of view of UKI. We adopt simple yet effective neural architectures and discuss the challenges of such an approach. Finally, we identify KE as a form of symbolic representation. From this perspective, we remark on the need of defining sophisticated UKI processes to verify the validity of knowledge integration. To this end, we foresee frameworks capable of combining symbolic and sub-symbolic representations for learning as a solution.

Flexible Computing Systems For AI Acceleration At The Extreme Edge Of The IoT

Embedding intelligence in extreme edge devices allows distilling raw data acquired from sensors into actionable information, directly on IoT end-nodes. This computing paradigm, in which end-nodes no longer depend entirely on the Cloud, offers undeniable benefits, driving a large research area (TinyML) to deploy leading Machine Learning (ML) algorithms on micro-controller class of devices. To fit the limited memory storage capability of these tiny platforms, full-precision Deep Neural Networks (DNNs) are compressed by representing their data down to byte and sub-byte formats, in the integer domain. However, the current generation of micro-controller systems can barely cope with the computing requirements of QNNs. This thesis tackles the challenge from many perspectives, presenting solutions both at software and hardware levels, exploiting parallelism, heterogeneity and software programmability to guarantee high flexibility and high energy-performance proportionality. The first contribution, PULP-NN, is an optimized software computing library for QNN inference on parallel ultra-low-power (PULP) clusters of RISC-V processors, showing one order of magnitude improvements in performance and energy efficiency, compared to current State-of-the-Art (SoA) STM32 micro-controller systems (MCUs) based on ARM Cortex-M cores. The second contribution is XpulpNN, a set of RISC-V domain specific instruction set architecture (ISA) extensions to deal with sub-byte integer arithmetic computation. The solution, including the ISA extensions and the micro-architecture to support them, achieves energy efficiency comparable with dedicated DNN accelerators and surpasses the efficiency of SoA ARM Cortex-M based MCUs, such as the low-end STM32M4 and the high-end STM32H7 devices, by up to three orders of magnitude. To overcome the Von Neumann bottleneck while guaranteeing the highest flexibility, the final contribution integrates an Analog In-Memory Computing accelerator into the PULP cluster, creating a fully programmable heterogeneous fabric that demonstrates end-to-end inference capabilities of SoA MobileNetV2 models, showing two orders of magnitude performance improvements over current SoA analog/digital solutions.

Interpretable Convolutional Neural Networks for Decoding and Analyzing Neural Time Series Data

Machine learning is widely adopted to decode multi-variate neural time series, including electroencephalographic (EEG) and single-cell recordings. Recent solutions based on deep learning (DL) outperformed traditional decoders by automatically extracting relevant discriminative features from raw or minimally pre-processed signals. Convolutional Neural Networks (CNNs) have been successfully applied to EEG and are the most common DL-based EEG decoders in the state-of-the-art (SOA). However, the current research is affected by some limitations. SOA CNNs for EEG decoding usually exploit deep and heavy structures with the risk of overfitting small datasets, and architectures are often defined empirically. Furthermore, CNNs are mainly validated by designing within-subject decoders. Crucially, the automatically learned features mainly remain unexplored; conversely, interpreting these features may be of great value to use decoders also as analysis tools, highlighting neural signatures underlying the different decoded brain or behavioral states in a data-driven way. Lastly, SOA DL-based algorithms used to decode single-cell recordings rely on more complex, slower to train and less interpretable networks than CNNs, and the use of CNNs with these signals has not been investigated. This PhD research addresses the previous limitations, with reference to P300 and motor decoding from EEG, and motor decoding from single-neuron activity. CNNs were designed light, compact, and interpretable. Moreover, multiple training strategies were adopted, including transfer learning, which could reduce training times promoting the application of CNNs in practice. Furthermore, CNN-based EEG analyses were proposed to study neural features in the spatial, temporal and frequency domains, and proved to better highlight and enhance relevant neural features related to P300 and motor states than canonical EEG analyses. Remarkably, these analyses could be used, in perspective, to design novel EEG biomarkers for neurological or neurodevelopmental disorders. Lastly, CNNs were developed to decode single-neuron activity, providing a better compromise between performance and model complexity.

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.

Applied deep learning and data science with a human-centric and data-centric approach

The term Artificial intelligence acquired a lot of baggage since its introduction and in its current incarnation is synonymous with Deep Learning. The sudden availability of data and computing resources has opened the gates to myriads of applications. Not all are created equal though, and problems might arise especially for fields not closely related to the tasks that pertain tech companies that spearheaded DL. The perspective of practitioners seems to be changing, however. Human-Centric AI emerged in the last few years as a new way of thinking DL and AI applications from the ground up, with a special attention at their relationship with humans. The goal is designing a system that can gracefully integrate in already established workflows, as in many real-world scenarios AI may not be good enough to completely replace its humans. Often this replacement may even be unneeded or undesirable. Another important perspective comes from, Andrew Ng, a DL pioneer, who recently started shifting the focus of development from “better models” towards better, and smaller, data. He defined his approach Data-Centric AI. Without downplaying the importance of pushing the state of the art in DL, we must recognize that if the goal is creating a tool for humans to use, more raw performance may not align with more utility for the final user. A Human-Centric approach is compatible with a Data-Centric one, and we find that the two overlap nicely when human expertise is used as the driving force behind data quality. This thesis documents a series of case-studies where these approaches were employed, to different extents, to guide the design and implementation of intelligent systems. We found human expertise proved crucial in improving datasets and models. The last chapter includes a slight deviation, with studies on the pandemic, still preserving the human and data centric perspective.

Governing algorithms in the big data era for balancing new digital rights - designing GDPR compliant and trustworthy XAI systems

This thesis investigates the legal, ethical, technical, and psychological issues of general data processing and artificial intelligence practices and the explainability of AI systems. It consists of two main parts. In the initial section, we provide a comprehensive overview of the big data processing ecosystem and the main challenges we face today. We then evaluate the GDPR’s data privacy framework in the European Union. The Trustworthy AI Framework proposed by the EU’s High-Level Expert Group on AI (AI HLEG) is examined in detail. The ethical principles for the foundation and realization of Trustworthy AI are analyzed along with the assessment list prepared by the AI HLEG. Then, we list the main big data challenges the European researchers and institutions identified and provide a literature review on the technical and organizational measures to address these challenges. A quantitative analysis is conducted on the identified big data challenges and the measures to address them, which leads to practical recommendations for better data processing and AI practices in the EU. In the subsequent part, we concentrate on the explainability of AI systems. We clarify the terminology and list the goals aimed at the explainability of AI systems. We identify the reasons for the explainability-accuracy trade-off and how we can address it. We conduct a comparative cognitive analysis between human reasoning and machine-generated explanations with the aim of understanding how explainable AI can contribute to human reasoning. We then focus on the technical and legal responses to remedy the explainability problem. In this part, GDPR’s right to explanation framework and safeguards are analyzed in-depth with their contribution to the realization of Trustworthy AI. Then, we analyze the explanation techniques applicable at different stages of machine learning and propose several recommendations in chronological order to develop GDPR-compliant and Trustworthy XAI systems.

Meaningful insights: explainability techniques for black-box models on tabular data

Artificial Intelligence (AI) and Machine Learning (ML) are novel data analysis techniques providing very accurate prediction results. They are widely adopted in a variety of industries to improve efficiency and decision-making, but they are also being used to develop intelligent systems. Their success grounds upon complex mathematical models, whose decisions and rationale are usually difficult to comprehend for human users to the point of being dubbed as black-boxes. This is particularly relevant in sensitive and highly regulated domains. To mitigate and possibly solve this issue, the Explainable AI (XAI) field became prominent in recent years. XAI consists of models and techniques to enable understanding of the intricated patterns discovered by black-box models. In this thesis, we consider model-agnostic XAI techniques, which can be applied to Tabular data, with a particular focus on the Credit Scoring domain. Special attention is dedicated to the LIME framework, for which we propose several modifications to the vanilla algorithm, in particular: a pair of complementary Stability Indices that accurately measure LIME stability, and the OptiLIME policy which helps the practitioner finding the proper balance among explanations' stability and reliability. We subsequently put forward GLEAMS a model-agnostic surrogate interpretable model which requires to be trained only once, while providing both Local and Global explanations of the black-box model. GLEAMS produces feature attributions and what-if scenarios, from both dataset and model perspective. Eventually, we argue that synthetic data are an emerging trend in AI, being more and more used to train complex models instead of original data. To be able to explain the outcomes of such models, we must guarantee that synthetic data are reliable enough to be able to translate their explanations to real-world individuals. To this end we propose DAISYnt, a suite of tests to measure synthetic tabular data quality and privacy.

Novel methods of state estimation and parameter identification for high-performance motorcycles

This research activity aims at providing a reliable estimation of particular state variables or parameters concerning the dynamics and performance optimization of a MotoGP-class motorcycle, integrating the classical model-based approach with new methodologies involving artificial intelligence. The first topic of the research focuses on the estimation of the thermal behavior of the MotoGP carbon braking system. Numerical tools are developed to assess the instantaneous surface temperature distribution in the motorcycle's front brake discs. Within this application other important brake parameters are identified using Kalman filters, such as the disc convection coefficient and the power distribution in the disc-pads contact region. Subsequently, a physical model of the brake is built to estimate the instantaneous braking torque. However, the results obtained with this approach are highly limited by the knowledge of the friction coefficient (μ) between the disc rotor and the pads. Since the value of μ is a highly nonlinear function of many variables (namely temperature, pressure and angular velocity of the disc), an analytical model for the friction coefficient estimation appears impractical to establish. To overcome this challenge, an innovative hybrid solution is implemented, combining the benefit of artificial intelligence (AI) with classical model-based approach. Indeed, the disc temperature estimated through the thermal model previously implemented is processed by a machine learning algorithm that outputs the actual value of the friction coefficient thus improving the braking torque computation performed by the physical model of the brake. Finally, the last topic of this research activity regards the development of an AI algorithm to estimate the current sideslip angle of the motorcycle's front tire. While a single-track motorcycle kinematic model and IMU accelerometer signals theoretically enable sideslip calculation, the presence of accelerometer noise leads to a significant drift over time. To address this issue, a long short-term memory (LSTM) network is implemented.

Identificazione e classificazione delle lesioni endocavitarie uterine dalle immagini isteroscopiche con il deep learning

Anche se l'isteroscopia con la biopsia endometriale è il gold standard nella diagnosi della patologia intracavitaria uterina, l'esperienza dell’isteroscopista è fondamentale per una diagnosi corretta. Il Deep Learning (DL) come metodica di intelligenza artificiale potrebbe essere un aiuto per superare questo limite. Sono disponibili pochi studi con risultati preliminari e mancano ricerche che valutano le prestazioni dei modelli di DL nell'identificazione delle lesioni intrauterine e il possibile aiuto derivato dai fattori clinici. Obiettivo: Sviluppare un modello di DL per identificare e classificare le patologie endocavitarie uterine dalle immagini isteroscopiche. Metodi: È stato eseguito uno studio di coorte retrospettivo osservazionale monocentrico su una serie consecutiva di casi isteroscopici di pazienti con patologia intracavitaria uterina confermata all’esame istologico eseguiti al Policlinico S. Orsola. Le immagini isteroscopiche sono state usate per costruire un modello di DL per la classificazione e l'identificazione delle lesioni intracavitarie con e senza l'aiuto di fattori clinici (età, menopausa, AUB, terapia ormonale e tamoxifene). Come risultati dello studio abbiamo calcolato le metriche diagnostiche del modello di DL nella classificazione e identificazione delle lesioni uterine intracavitarie con e senza l'aiuto dei fattori clinici. Risultati: Abbiamo esaminato 1.500 immagini provenienti da 266 casi: 186 pazienti avevano lesioni focali benigne, 25 lesioni diffuse benigne e 55 lesioni preneoplastiche/neoplastiche. Sia per quanto riguarda la classificazione che l’identificazione, le migliori prestazioni sono state raggiunte con l'aiuto dei fattori clinici, complessivamente con precision dell'80,11%, recall dell'80,11%, specificità del 90,06%, F1 score dell’80,11% e accuratezza dell’86,74% per la classificazione. Per l’identificazione abbiamo ottenuto un rilevamento complessivo dell’85,82%, precision 93,12%, recall del 91,63% ed F1 score del 92,37%. Conclusioni: Il modello DL ha ottenuto una bassa performance nell’identificazione e classificazione delle lesioni intracavitarie uterine dalle immagini isteroscopiche. Anche se la migliore performance diagnostica è stata ottenuta con l’aiuto di fattori clinici specifici, questo miglioramento è stato scarso.

High-resolution stratigraphy and seismic site characterization of late Quaternary paleovalley systems (Adriatic coastal plain, Italy)

In highly urbanized coastal lowlands, effective site characterization is crucial for assessing seismic risk. It requires a comprehensive stratigraphic analysis of the shallow subsurface, coupled with the precise assessment of the geophysical properties of buried deposits. In this context, late Quaternary paleovalley systems, shallowly buried fluvial incisions formed during the Late Pleistocene sea-level fall and filled during the Holocene sea-level rise, are crucial for understanding seismic amplification due to their soft sediment infill and sharp lithologic contrasts. In this research, we conducted high-resolution stratigraphic analyses of two regions, the Pescara and Manfredonia areas along the Adriatic coastline of Italy, to delineate the geometries and facies architecture of two paleovalley systems. Furthermore, we carried out geophysical investigations to characterize the study areas and perform seismic response analyses. We tested the microtremor-based horizontal-to-vertical spectral ratio as a mapping tool to reconstruct the buried paleovalley geometries. We evaluated the relationship between geological and geophysical data and identified the stratigraphic surfaces responsible for the observed resonances. To perform seismic response analysis of the Pescara paleovalley system, we integrated the stratigraphic framework with microtremor and shear wave velocity measurements. The seismic response analysis highlights strong seismic amplifications in frequency ranges that can interact with a wide variety of building types. Additionally, we explored the applicability of artificial intelligence in performing facies analysis from borehole images. We used a robust dataset of high-resolution digital images from continuous sediment cores of Holocene age to outline a novel, deep-learning-based approach for performing automatic semantic segmentation directly on core images, leveraging the power of convolutional neural networks. We propose an automated model to rapidly characterize sediment cores, reproducing the sedimentologist's interpretation, and providing guidance for stratigraphic correlation and subsurface reconstructions.

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.