A model and an algebra for semi-structured and full-text queries

The need for a convergence between semi-structured data management and Information Retrieval techniques is manifest to the scientific community. In order to fulfil this growing request, W3C has recently proposed XQuery Full Text, an IR-oriented extension of XQuery. However, the issue of query optimization requires the study of important properties like query equivalence and containment; to this aim, a formal representation of document and queries is needed. The goal of this thesis is to establish such formal background. We define a data model for XML documents and propose an algebra able to represent most of XQuery Full-Text expressions. We show how an XQuery Full-Text expression can be translated into an algebraic expression and how an algebraic expression can be optimized.

Kernel Methods for Tree Structured Data

Machine learning comprises a series of techniques for automatic extraction of meaningful information from large collections of noisy data. In many real world applications, data is naturally represented in structured form. Since traditional methods in machine learning deal with vectorial information, they require an a priori form of preprocessing. Among all the learning techniques for dealing with structured data, kernel methods are recognized to have a strong theoretical background and to be effective approaches. They do not require an explicit vectorial representation of the data in terms of features, but rely on a measure of similarity between any pair of objects of a domain, the kernel function. Designing fast and good kernel functions is a challenging problem. In the case of tree structured data two issues become relevant: kernel for trees should not be sparse and should be fast to compute. The sparsity problem arises when, given a dataset and a kernel function, most structures of the dataset are completely dissimilar to one another. In those cases the classifier has too few information for making correct predictions on unseen data. In fact, it tends to produce a discriminating function behaving as the nearest neighbour rule. Sparsity is likely to arise for some standard tree kernel functions, such as the subtree and subset tree kernel, when they are applied to datasets with node labels belonging to a large domain. A second drawback of using tree kernels is the time complexity required both in learning and classification phases. Such a complexity can sometimes prevents the kernel application in scenarios involving large amount of data. This thesis proposes three contributions for resolving the above issues of kernel for trees. A first contribution aims at creating kernel functions which adapt to the statistical properties of the dataset, thus reducing its sparsity with respect to traditional tree kernel functions. Specifically, we propose to encode the input trees by an algorithm able to project the data onto a lower dimensional space with the property that similar structures are mapped similarly. By building kernel functions on the lower dimensional representation, we are able to perform inexact matchings between different inputs in the original space. A second contribution is the proposal of a novel kernel function based on the convolution kernel framework. Convolution kernel measures the similarity of two objects in terms of the similarities of their subparts. Most convolution kernels are based on counting the number of shared substructures, partially discarding information about their position in the original structure. The kernel function we propose is, instead, especially focused on this aspect. A third contribution is devoted at reducing the computational burden related to the calculation of a kernel function between a tree and a forest of trees, which is a typical operation in the classification phase and, for some algorithms, also in the learning phase. We propose a general methodology applicable to convolution kernels. Moreover, we show an instantiation of our technique when kernels such as the subtree and subset tree kernels are employed. In those cases, Direct Acyclic Graphs can be used to compactly represent shared substructures in different trees, thus reducing the computational burden and storage requirements.

Electrosynthesis and characterization of structured catalysts

In this research work the optimization of the electrochemical system of LDHs as catalytic precursors on FeCrAlY foams was carried out. Preliminary sintheses were performed on flat surfaces in order to easily characterize the deposited material. From the study of pH evolution vs time at different cathodic potentials applied to a Pt electrode, the theoretical best working conditions for the synthesis of single hydroxides and LDH compounds was achieved. In order to define the optimal potential for the synthesis of a particular LDH compound, the collected data were compared with the interval of precipitation determined by titration with NaOH. However, the characterization of the deposited material on Pt surfaces did not confirm the deposition of a pure and homogeneous LDH phase during the synthesis. Instead a sequential deposition linked to the pH of precipitation of the involved elements is observed. The same behavior was observed during the synthesis of the RhMgAl LDH on FeCrAlY foam as catalytic precursor. Several parameters were considered in order to optimize the synthesis.. The development of electrochemical cells with different feature, such as the counter electrode dimensions or the contact between the foam and the potentiostat, had been carried out in order to obtain a better coating of the foam. The influence of the initial pH of the electrolyte solution, of the applied potential, of the composition of the electrolytic solution were investigated in order to improve a better coating of the catalyst support. Catalytic tests were performed after the calcination of the deposited foam for the CPO and SR reactions, showing an improve of performances along with optimization of the precursors synthesis conditions.

Development of newly conceived biomimetic nano-structured biomaterials as scaffolds for bone and osteochondral regeneration

The present research thesis was focused on the development of new biomaterials and devices for application in regenerative medicine, particularly in the repair/regeneration of bone and osteochondral regions affected by degenerative diseases such as Osteoarthritis and Osteoporosis or serious traumas. More specifically, the work was focused on the synthesis and physico-chemical-morphological characterization of: i) a new superparamagnetic apatite phase; ii) new biomimetic superparamagnetic bone and osteochondral scaffolds; iii) new bioactive bone cements for regenerative vertebroplasty. The new bio-devices were designed to exhibit high biomimicry with hard human tissues and with functionality promoting faster tissue repair and improved texturing. In particular, recent trends in tissue regeneration indicate magnetism as a new tool to stimulate cells towards tissue formation and organization; in this perspective a new superparamagnetic apatite was synthesized by doping apatite lattice with di-and trivalent iron ions during synthesis. This finding was the pin to synthesize newly conceived superparamagnetic bone and osteochondral scaffolds by reproducing in laboratory the biological processes yielding the formation of new bone, i.e. the self-assembly/organization of collagen fibrils and heterogeneous nucleation of nanosized, ionically substituted apatite mimicking the mineral part of bone. The new scaffolds can be magnetically switched on/off and function as workstations guiding fast tissue regeneration by minimally invasive and more efficient approaches. Moreover, in the view of specific treatments for patients affected by osteoporosis or traumas involving vertebrae weakening or fracture, the present work was also dedicated to the development of new self-setting injectable pastes based on strontium-substituted calcium phosphates, able to harden in vivo and transform into strontium-substituted hydroxyapatite. The addition of strontium may provide an anti-osteoporotic effect, aiding to restore the physiologic bone turnover. The ceramic-based paste was also added with bio-polymers, able to be progressively resorbed thus creating additional porosity in the cement body that favour cell colonization and osseointegration.

Nanofibrous-structured polymers and their engineering applications

Monolithic materials cannot always satisfy the demands of today’s advanced requirements. Only by combining several materials at different length-scales, as nature does, the requested performances can be met. Polymer nanocomposites are intended to overcome the common drawbacks of pristine polymers, with a multidisciplinary collaboration of material science with chemistry, engineering, and nanotechnology. These materials are an active combination of polymers and nanomaterials, where at least one phase lies in the nanometer range. By mimicking nature’s materials is possible to develop new nanocomposites for structural applications demanding combinations of strength and toughness. In this perspective, nanofibers obtained by electrospinning have been increasingly adopted in the last decade to improve the fracture toughness of Fiber Reinforced Plastic (FRP) laminates. Although nanofibers have already found applications in various fields, their widespread introduction in the industrial context is still a long way to go. This thesis aims to develop methodologies and models able to predict the behaviour of nanofibrous-reinforced polymers, paving the way for their practical engineering applications. It consists of two main parts. The first one investigates the mechanisms that act at the nanoscale, systematically evaluating the mechanical properties of both the nanofibrous reinforcement phase (Chapter 1) and hosting polymeric matrix (Chapter 2). The second part deals with the implementation of different types of nanofibers for novel pioneering applications, trying to combine the well-known fracture toughness enhancement in composite laminates with improving other mechanical properties or including novel functionalities. Chapter 3 reports the development of novel adhesive carriers made of nylon 6,6 nanofibrous mats to increase the fracture toughness of epoxy-bonded joints. In Chapter 4, recently developed rubbery nanofibers are used to enhance the damping properties of unidirectional carbon fiber laminates. Lastly, in Chapter 5, a novel self-sensing composite laminate capable of detecting impacts on its surface using PVDF-TrFE piezoelectric nanofibers is presented.

Learning representations for graph-structured socio-technical systems

The recent widespread use of social media platforms and web services has led to a vast amount of behavioral data that can be used to model socio-technical systems. A significant part of this data can be represented as graphs or networks, which have become the prevalent mathematical framework for studying the structure and the dynamics of complex interacting systems. However, analyzing and understanding these data presents new challenges due to their increasing complexity and diversity. For instance, the characterization of real-world networks includes the need of accounting for their temporal dimension, together with incorporating higher-order interactions beyond the traditional pairwise formalism. The ongoing growth of AI has led to the integration of traditional graph mining techniques with representation learning and low-dimensional embeddings of networks to address current challenges. These methods capture the underlying similarities and geometry of graph-shaped data, generating latent representations that enable the resolution of various tasks, such as link prediction, node classification, and graph clustering. As these techniques gain popularity, there is even a growing concern about their responsible use. In particular, there has been an increased emphasis on addressing the limitations of interpretability in graph representation learning. This thesis contributes to the advancement of knowledge in the field of graph representation learning and has potential applications in a wide range of complex systems domains. We initially focus on forecasting problems related to face-to-face contact networks with time-varying graph embeddings. Then, we study hyperedge prediction and reconstruction with simplicial complex embeddings. Finally, we analyze the problem of interpreting latent dimensions in node embeddings for graphs. The proposed models are extensively evaluated in multiple experimental settings and the results demonstrate their effectiveness and reliability, achieving state-of-the-art performances and providing valuable insights into the properties of the learned representations.

Towards a sustainable approach for Cultural Heritage Conservation: developing micro-structured green materials and analytical protocols for their performance assessment

Cleaning is one of the most important and delicate procedures that are part of the restoration process. When developing new systems, it is fundamental to consider its selectivity towards the layer to-be-removed, non-invasiveness towards the one to-be-preserved, its sustainability and non-toxicity. Besides assessing its efficacy, it is important to understand its mechanism by analytical protocols that strike a balance between cost, practicality, and reliable interpretation of results. In this thesis, the development of cleaning systems based on the coupling of electrospun fabrics (ES) and greener organic solvents is proposed. Electrospinning is a versatile technique that allows the production of micro/nanostructured non-woven mats, which have already been used as absorbents in various scientific fields, but to date, not in the restoration field. The systems produced proved to be effective for the removal of dammar varnish from paintings, where the ES not only act as solvent-binding agents but also as adsorbents towards the partially solubilised varnish due to capillary rise, thus enabling a one-step procedure. They have also been successfully applied for the removal of spray varnish from marble substrates and wall paintings. Due to the materials' complexity, the procedure had to be adapted case-by-case and mechanical action was still necessary. According to the spinning solution, three types of ES mats have been produced: polyamide 6,6, pullulan and pullulan with melanin nanoparticles. The latter, under irradiation, allows for a localised temperature increase accelerating and facilitating the removal of less soluble layers (e.g. reticulated alkyd-based paints). All the systems produced, and the mock-ups used were extensively characterised using multi-analytical protocols. Finally, a monitoring protocol and image treatment based on photoluminescence macro-imaging is proposed. This set-up allowed the study of the removal mechanism of dammar varnish and semi-quantify its residues. These initial results form the basis for optimising the acquisition set-up and data processing.