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.

Use of computer algebra in the calculation of Feynman diagrams

The increasing precision of current and future experiments in high-energy physics requires a likewise increase in the accuracy of the calculation of theoretical predictions, in order to find evidence for possible deviations of the generally accepted Standard Model of elementary particles and interactions. Calculating the experimentally measurable cross sections of scattering and decay processes to a higher accuracy directly translates into including higher order radiative corrections in the calculation. The large number of particles and interactions in the full Standard Model results in an exponentially growing number of Feynman diagrams contributing to any given process in higher orders. Additionally, the appearance of multiple independent mass scales makes even the calculation of single diagrams non-trivial. For over two decades now, the only way to cope with these issues has been to rely on the assistance of computers. The aim of the xloops project is to provide the necessary tools to automate the calculation procedures as far as possible, including the generation of the contributing diagrams and the evaluation of the resulting Feynman integrals. The latter is based on the techniques developed in Mainz for solving one- and two-loop diagrams in a general and systematic way using parallel/orthogonal space methods. These techniques involve a considerable amount of symbolic computations. During the development of xloops it was found that conventional computer algebra systems were not a suitable implementation environment. For this reason, a new system called GiNaC has been created, which allows the development of large-scale symbolic applications in an object-oriented fashion within the C++ programming language. This system, which is now also in use for other projects besides xloops, is the main focus of this thesis. The implementation of GiNaC as a C++ library sets it apart from other algebraic systems. Our results prove that a highly efficient symbolic manipulator can be designed in an object-oriented way, and that having a very fine granularity of objects is also feasible. The xloops-related parts of this work consist of a new implementation, based on GiNaC, of functions for calculating one-loop Feynman integrals that already existed in the original xloops program, as well as the addition of supplementary modules belonging to the interface between the library of integral functions and the diagram generator.

Le matrici e la loro algebra

In questa tesi ci si propone lo studio dell'anello delle matrici quadrate di ordine n, su un campo, per arrivare a dimostrare che ha solo ideali banali pur non essendo un campo. Allo scopo si introducono le operazioni elementari e il procedimento di traduzione di tali operazioni con opportune moltiplicazioni per matrici dette elementari. Si considera inoltre il gruppo generale lineare arrivando a dimostrare che un particolare sottoinsieme delle matrici elementari è un generatore di tale gruppo.

Segmentazione di immagini mammografiche con convolutional neural networks

Il tumore al seno si colloca al primo posto per livello di mortalità tra le patologie tumorali che colpiscono la popolazione femminile mondiale. Diversi studi clinici hanno dimostrato come la diagnosi da parte del radiologo possa essere aiutata e migliorata dai sistemi di Computer Aided Detection (CAD). A causa della grande variabilità di forma e dimensioni delle masse tumorali e della somiglianza di queste con i tessuti che le ospitano, la loro ricerca automatizzata è un problema estremamente complicato. Un sistema di CAD è generalmente composto da due livelli di classificazione: la detection, responsabile dell’individuazione delle regioni sospette presenti sul mammogramma (ROI) e quindi dell’eliminazione preventiva delle zone non a rischio; la classificazione vera e propria (classification) delle ROI in masse e tessuto sano. Lo scopo principale di questa tesi è lo studio di nuove metodologie di detection che possano migliorare le prestazioni ottenute con le tecniche tradizionali. Si considera la detection come un problema di apprendimento supervisionato e lo si affronta mediante le Convolutional Neural Networks (CNN), un algoritmo appartenente al deep learning, nuova branca del machine learning. Le CNN si ispirano alle scoperte di Hubel e Wiesel riguardanti due tipi base di cellule identificate nella corteccia visiva dei gatti: le cellule semplici (S), che rispondono a stimoli simili ai bordi, e le cellule complesse (C) che sono localmente invarianti all’esatta posizione dello stimolo. In analogia con la corteccia visiva, le CNN utilizzano un’architettura profonda caratterizzata da strati che eseguono sulle immagini, alternativamente, operazioni di convoluzione e subsampling. Le CNN, che hanno un input bidimensionale, vengono solitamente usate per problemi di classificazione e riconoscimento automatico di immagini quali oggetti, facce e loghi o per l’analisi di documenti.

Algebra esterna e grassmanniane

Nella tesi viene fornita una costruzione dell'algebra esterna di un K-spazio vettoriale, alcune conseguenze principali come la derivazione in maniera traspente del determinante di e alcune sue proprietà e l'introduzione del concetto di Grassmanniana.

Deep learning for computer vision: a comparison between convolutional neural networks and hierarchical temporal memories on object recognition tasks

In recent years, Deep Learning techniques have shown to perform well on a large variety of problems both in Computer Vision and Natural Language Processing, reaching and often surpassing the state of the art on many tasks. The rise of deep learning is also revolutionizing the entire field of Machine Learning and Pattern Recognition pushing forward the concepts of automatic feature extraction and unsupervised learning in general. However, despite the strong success both in science and business, deep learning has its own limitations. It is often questioned if such techniques are only some kind of brute-force statistical approaches and if they can only work in the context of High Performance Computing with tons of data. Another important question is whether they are really biologically inspired, as claimed in certain cases, and if they can scale well in terms of "intelligence". The dissertation is focused on trying to answer these key questions in the context of Computer Vision and, in particular, Object Recognition, a task that has been heavily revolutionized by recent advances in the field. Practically speaking, these answers are based on an exhaustive comparison between two, very different, deep learning techniques on the aforementioned task: Convolutional Neural Network (CNN) and Hierarchical Temporal memory (HTM). They stand for two different approaches and points of view within the big hat of deep learning and are the best choices to understand and point out strengths and weaknesses of each of them. CNN is considered one of the most classic and powerful supervised methods used today in machine learning and pattern recognition, especially in object recognition. CNNs are well received and accepted by the scientific community and are already deployed in large corporation like Google and Facebook for solving face recognition and image auto-tagging problems. HTM, on the other hand, is known as a new emerging paradigm and a new meanly-unsupervised method, that is more biologically inspired. It tries to gain more insights from the computational neuroscience community in order to incorporate concepts like time, context and attention during the learning process which are typical of the human brain. In the end, the thesis is supposed to prove that in certain cases, with a lower quantity of data, HTM can outperform CNN.

Il polinomio strutturale di un'algebra monounaria

Questa tesi descrive alcune proprietà delle algebre monounarie finite e si propone di trovare un metodo per classificarle. Poiché infatti il numero di algebre di ordine n aumenta notevolmente con la crescita di quest’ultimo, si cerca un modo per suddividerle in classi d’isomorfismo. In particolare, dal momento che anche il numero di queste classi cresce esponenzialmente all’aumentare di n, utilizziamo una classificazione meno fine dell’isomorfismo basata sul polinomio strutturale. Grazie a questo strumento infatti è possibile risalire a famiglie di grafi orientati associati ad algebre monounarie, a due a due non isomorfi, ricavando perciò alcune specifiche caratteristiche di quest’ultime. Infine, calcolando l’ordine di gruppi particolari, detti automorfi, si può ottenere l’effettivo numero di algebre aventi un dato polinomio strutturale.

Computer algebra for x-ray spectral reconstruction between 6 and 25 MV

A previously presented algorithm for the reconstruction of bremsstrahlung spectra from transmission data has been implemented into MATHEMATICA. Spectra vectorial algebra has been used to solve the matrix system A * F = T. The new implementation has been tested by reconstructing photon spectra from transmission data acquired in narrow beam conditions, for nominal energies of 6, 15, and 25 MV. The results were in excellent agreement with the original calculations. Our implementation has the advantage to be based on a well-tested mathematical kernel. Furthermore it offers a comfortable user interface.

A feasible theory of truth over combinatory algebra

We define an applicative theory of truth TPT which proves totality exactly for the polynomial time computable functions. TPT has natural and simple axioms since nearly all its truth axioms are standard for truth theories over an applicative framework. The only exception is the axiom dealing with the word predicate. The truth predicate can only reflect elementhood in the words for terms that have smaller length than a given word. This makes it possible to achieve the very low proof-theoretic strength. Truth induction can be allowed without any constraints. For these reasons the system TPT has the high expressive power one expects from truth theories. It allows embeddings of feasible systems of explicit mathematics and bounded arithmetic. The proof that the theory TPT is feasible is not easy. It is not possible to apply a standard realisation approach. For this reason we develop a new realisation approach whose realisation functions work on directed acyclic graphs. In this way, we can express and manipulate realisation information more efficiently.

Algebra of Classical Theoretical Term Reductions in the Sciences

An elementary algebra identifies conceptual and corresponding applicational limitations in John Kemeny and Paul Oppenheim’s (K-O) 1956 model of theoretical reduction in the sciences. The K-O model was once widely accepted, at least in spirit, but seems afterward to have been discredited, or in any event superceeded. Today, the K-O reduction model is seldom mentioned, except to clarify when a reduction in the Kemeny-Oppenheim sense is not intended. The present essay takes a fresh look at the basic mathematics of K-O comparative vocabulary theoretical term reductions, from historical and philosophical standpoints, as a contribution to the history of the philosophy of science. The K-O theoretical reduction model qualifies a theory replacement as a successful reduction when preconditions of explanatory adequacy and comparable systematicization are met, and there occur fewer numbers of theoretical terms identified as replicable syntax types in the most economical statement of a theory’s putative propositional truths, as compared with the theoretical term count for the theory it replaces. The challenge to the historical model developed here, to help explain its scope and limitations, involves the potential for equivocal theoretical meanings of multiple theoretical term tokens of the same syntactical type.

Food Recognition for Dietary Assessment Using Deep Convolutional Neural Networks

Diet management is a key factor for the prevention and treatment of diet-related chronic diseases. Computer vision systems aim to provide automated food intake assessment using meal images. We propose a method for the recognition of already segmented food items in meal images. The method uses a 6-layer deep convolutional neural network to classify food image patches. For each food item, overlapping patches are extracted and classified and the class with the majority of votes is assigned to it. Experiments on a manually annotated dataset with 573 food items justified the choice of the involved components and proved the effectiveness of the proposed system yielding an overall accuracy of 84.9%.