409 research outputs found
Sort by
Automated Learning of Quantitative Software Models from System Traces
Full text linkModels are primary artifacts in software system development.
In particular, performance models allow us to evaluate and
reason about extra-functional properties, such as the aver-
age response time and throughput, for which meeting ade-
quate quality levels is increasingly important. Indeed, per-
formance quality is considered as essential as correctness in many practical development scenarios. Markov processes are valuable models for the qualitative analysis of performance.
In this thesis, we will present statistical methods that learn Markov models directly from the running software system traces. We will focus on two classes of processes: with and without memory. In the first scenario, we aim to learn funda-mental performance metrics, i.e., service demands and rout-ing probabilities, using queuing networks (QN). For processes with memory, instead, we will exploit variable length Markov chains (VLMC) to capture data dependencies throughout the traces of system executions. The conducted numerical eval-uations, the presented in-depth study of the literature, and the performed appropriate comparisons with similar tools al-low us to demonstrate how the approaches presented in this work constitute a significant step forward concerning state of the art
Shipwrecking Probability in Mediterranean Territorial Waters. A Cultural Approach to Archaeological Predictive Modelling
Full text linkThis thesis presents a formal approach and a GIS-based methodology for the assessment of the
shipwrecking probability in Mediterranean territorial waters, thus addressing the
underdevelopment of archaeological predictive models in the maritime domain, particularly in the
Mediterranean region. As archaeological predictive models are often criticized for oversimplifying
complex historical phenomena to produce quantifiable outcomes, this study focuses on two
different scales of analysis to meet the need for both a general tool applicable to spatial planning
and a more detailed one providing insights for historical and archaeological research. First, a
regional-scale model is developed, which focuses on navigation dynamics in the area between Cap
Bon (present Tunisia) and Alexandria (present Egypt) in Roman times. Then, this model is
extended to all Mediterranean territorial waters in a simplified version and without chronological
limitations. At both scales, the criteria for selecting the input factors are formalized.
In order to identify areas with higher shipwrecking probability than others, two sub-questions are
addressed that correspond to separate model components: 1. Where would ships be more likely to
transit? 2. Where would ships have a higher risk of sinking? Grounding the theory-building on a
systematic screening of accounts by primary sources, the first model component derives transit
probabilities by considering multiple, oftentimes competing, criteria that trigger and affect
mariners’ movements, including in particular the effects of risk perception - thus rejecting the idea
that sailors would necessarily choose the optimal or most efficient route. The second model
component includes environmental hazards objectively increasing the risk of sinking.
Given the many elements of uncertainty and subjective reasoning behind the model building - a
problem often unheeded in archaeological computational modelling - an entire chapter is devoted
to a sensitivity analysis of the model and the exploration of diverse model scenarios. The overall
methodology attempts to overcome some of the main pitfalls of current modelling approaches to
seafaring and to shipwreck locations, namely, the inductive use of shipwreck data without a formal
exploration of data biases, and the predominant reliance on environmental and economic input
variables to the detriment of cultural and cognitive factors.
This study suggests that by explicitly differentiating between actual and perceived risks, and
accounting for the effects this difference produces in terms of variations from the optimal
navigation corridors, the predictive ability of the model increases. While constituting a valuable
tool for optimizing maritime spatial planning and archaeological investigations, this model also
offers insights into the biases in current shipwreck data. The model furthermore provides an
adaptable toolkit applicable to other geographical contexts and chronological periods, and a
suitable basis for expansion with a future component by modelling post-depositional dynamics
that affect the preservation and detectability of wrecks at local scales
Memorabilia pompeiana (1748-1830). Antiquities from Pompeii in the European collections
Full text linkBetween the 18th and 19th centuries, Naples and its kingdom were a lively hub for the antiquarian milieu both local and non-local. The Bourbon Crown made a noteworthy effort to safeguard the archaeological heritage of the State, especially following the beginning of the excavations in the Vesuvian sites: Herculaneum in 1738, Pompeii in 1748, and Stabia in 1749. In fact, these “regi scavi” were an allodial asset, meaning that they were a private property of the Bourbons.
Focusing on Pompeii, its (re)discovery in 1748 had an immediate echo in the cultural background at the time, and was widely spread throughout Europe. The extraordinary state of conservation of the artefacts recovered at Pompeii lured collectors from everywhere, who tried (successfully or not) to obtain some of these findings, despite the strict governmental regulations.
The main aim of this work is tracking down the Pompeiana exported to European collections both public and private, during the span time 1748-1830, namely from the debut of the Bourbon investigations in the ancient Campanian city to the death of King Francis I, thus playing out almost in parallel to the evolution of the legal system governing the historical-artistic heritage of the Neapolitan kingdom. Another important objective is to define the modes that allowed collectors to enter into possession of such a kind of antiquities. For each artefact taken into account, retracing its collecting itinerary is a fundamental goal; furthermore, identifying its contextual data is an intriguing challenge. A meticulous archival research, mainly in the Archivio di Stato di Napoli and the Archivio Storico del Museo Archeologico Nazionale di Napoli, has been essential to carry out the whole study
The power of victory: Italy, Austria and the struggle for treasures of art and history after the First World War (1918-1923)
Full text linkStemming from the author’s previous years of research into how peace settlements have changed the destiny of so many artistic treasures and entire
historical collections in Europe, this work appears to have been only a matter of time. Despite there being a striking amount of records and first-hand accounts on artistic disputes between Austria and Italy at the end of the First World War, recent contributions, albeit precious, have so far remained quite circumscribed,
often focused on either Italian or Austrian sources and perspectives, but seldom taking both into account in equal measure. The fact that no broader analysis had been undertaken yet is partly attributable to the perhaps less appealing and less fictional features of post-1918 claims and restitutions. Exactly because no big-scale, thoroughly organised looting campaigns directed by incredibly power-thirsty individuals preceded those events, it may have seemed somehow intimidating comparing them to the sensational retrievals of the Napoleonic loots in Paris in 1815 and the equally unprecedented and
gigantic effort of the Allies, and the Americans in particular, towards the recovery and restitution of artworks displaced by Nazi and Fascist authorities.
Yet, what happened less than thirty years before that is in a sense the unintentionally neglected link in a chain that seems worth appreciating in its entirety.
During the Congress of Vienna no official treaty ever acknowledged the restitutions extolled from France through the military intervention of Prussia and Austria and the mediation of the British in favour of Canova’s requests. For reasons of international tact, relations with the restored French monarchy were not to be publicly compromised from the very onset. Conversely, the authority of the peace agreements and official restitution policies was to constitute the linchpin of post-1945 restitutions as administered mainly by the United States and their Army officials, particularly keen on abidance by the protocol and well-defined, ideally universal procedures. In 1919 and the years that followed, one interestingly witnesses a sort of liminal situation that borrows from previous instances of forced retrievals at the expenses of the vanquished but at
the same time paves the way for a more regulated implementation of
restitution demands through peace treaties, official protocols and bilateral
agreements. This type of legal primary sources just mentioned, along with official and personal correspondence kept in so many archives, Italian but in particular Austrian newspaper articles and the accounts penned and published
by all those who had a direct role in the events, constitute not only the heart of this work, but its very reason to be.
Acknowledging the potential of such a deep but partly untapped pool of information, this research has thus been intentionally and fundamentally archive-based. As it will become clear throughout the main text, the account
has given a significant priority to primary sources over second-hand and late contributions. The main reason for that is the fact that many recent studies rely on and constitute an interpretation only of part of those records. A work built too much on recent literature would have not left enough room and silence for the original voices to be heard, appreciated and contextualised in the historical events that framed and influenced them. The choice of proceeding along the lines drawn by the available yet greatly unpublished material has furthermore highlighted the need for a comprehensive chronological account of events only
partially known to scholars and the public alike. Against the backdrop of those five years that, after 1918, slowly and painfully dragged collapsed empires,
vanquished and victorious nations out of the cruelties of the war, the work
traces the steps of those in Italy and Austria who kept fighting for objects and collections of art and history with undiminished urge. In what can be seen as a backlash of the real hostilities that had just subsumed, the presence of the Italian military in Vienna ushered in a period of recriminations and threats that resulted in the forced seizure of dozens of paintings and manuscripts from the major institutions of the Austrian capital. The reaction of local intellectuals and
the public opinion contributed to making it an international affair with serious
repercussions on the peace negotiations in Paris. And in this sense the upcoming treaties, agreements and the official directives of international bodies like the Reparation Commission ended up playing a paramount role in the destiny of national collections like the Austrian ones, under the persisting
threat of claimant countries for years after the war. What ensues is thus the analysis of how pressing Italian demands had to translate into more
accommodating and diplomatic attitudes, despite a race against time to avoid the entanglements and caveats of post-war diplomacy and regulations. In the end, the ultimate destiny of major public collections and unique objects of art and history had, like in the past, to be subjected to exceptional and unprecedented political circumstances and power struggles that more often than not go unnoticed in the general art-historical discourse.
After discovering how much had been left untold that was actually available through so many documents and writings on both sides of the Alps, the urgency to catch up and put together a comprehensive, transnational history of those years arose naturally. For this cannot be but a story told from the Austrian and the Italian side at the same time, especially after more than a century has passed. Such a multi-centred way of proceeding resorts to a wide,
almost infinite range of connections between people, objects and places and thus automatically transcends political boundaries. In so doing, it also advocates an interpretation of the facts that wants to be as little biased as possible, an interpretation that won’t intrude too much throughout the narration so as to let the reader appreciate first and foremost the events as they followed and triggered one another, leaving some food for thought only at the very end. Consistently, the choice was also that of trying not to fall into mainstream discourses of art looting and restitution. Comparisons with earlier and later examples will inevitably be drawn, but this specific chapter of the past and its characters will still retain their own historical dignity and
autonomy. This automatically entails leaving behind binary interpretations along the lines of good and wrong, of customary and unlawful, compensation and punishment, both on an individual and on a collective level. Paradoxically
though, what is provided here is no real alternative to existing narratives except a fresh look at something that still remains hopelessly complex but, for this very reason, ever enriching
Semi-supervised and weakly-supervised learning with spatio-temporal priors in medical image segmentation
Full text linkOver the last decades, medical imaging techniques have
played a crucial role in healthcare, supporting radiologists
and facilitating patient diagnosis. With the advent of faster
and higher-quality imaging technologies, the amount of data
that is possible to collect for each patient is paving the way
toward personalised medicine. As a result, automating simple
image analysis operations, such as lesion localisation and
quantification, would greatly help clinicians focus energy
and attention on tasks best done by human intelligence.
Most recently, Artificial Intelligence (AI) research is accelerating
in healthcare, providing tools that often perform on par or
even better than humans in conceptually simple image processing
operations. In our work, we pay special attention to
the problem of automating semantic segmentation, where an
image is partitioned into multiple semantically meaningful
regions, separating the anatomical components of interest.
Unfortunately, developing effective AI segmentation tools usually
needs large quantities of annotated data. Conversely,
obtaining large-scale annotated datasets is difficult in medical
imaging, as it requires experts and is time-consuming.
For this reason, we develop automated methods to reduce the
need for collecting high-quality annotated data, both in terms
of the number and type of required annotations. We make
this possible by constraining the data representation learned
by our method to be semantic or by regularising the model
predictions to satisfy data-driven spatio-temporal priors. In
the thesis, we also open new avenues for future research using
AI with limited annotations, which we believe is key to
developing robust AI models for medical image analysis
Machine learning methods for control, identification, and estimation
Full text linkOver the last decades, the landscape of control theory and
system identification has changed significantly in response
to the new challenges arising from the industry. This is not
surprising: classical model-based techniques are not suitable
to handle real-world applications for which it is often too
expensive to derive even an approximate model using first
principles. Data-driven approaches represent a solution to
such an issue. Thanks to the ever-increasing availability of a
large quantity of data, they have quickly become central topics
within the control theory community.
This thesis collects some results regarding using machine
learning approaches to answer some open questions in control
theory by formulating novel techniques and lessening
some undesirable aspects of existing methods. We first
present a system identification approach based on deep
learning to learn state-space models for nonlinear systems.
We then propose a data-driven virtual sensor synthesis approach,
inspired by the Multiple Model Adaptive Estimation
framework, for reconstructing normally unmeasurable quantities
such as scheduling parameters in parameter-varying
systems. Three data-driven control approaches, two of which
are based on the well-known Virtual Reference Feedback
Tuning framework, are finally presented to synthesize constrained
controllers for unknown nonlinear dynamical systems
from the data without identifying first a model of the
plant. Tuning guidelines for the proposed methods are also
provided
The effect of compression and expansion on stochastic reaction networks
Full text linkMarkov chains are a fundamental model to study systems
with stochastic behavior. However, their state space is often
of an unmanageable size, making the use of approximations
and simplifications necessary for analytic solutions. This thesis
considers reaction networks as a well-known representation
for Markov chains describing interactions between species
populations. It presents several methods using model transformations
to aid with the effective analysis of such systems.
Species equivalence is a reduction technique that lifts the concept
(and related algorithms) of Markov chain lumpability
from lumping of states to directly lumping species in a reaction
network. This allows the simplification of a reaction
network without first examining its state space.
The tool DiffLQN implements a method for the analysis of
large-scale stochastic models for the performance evaluation
of software systems using an approach based on deterministic
rate equations, by means of a compact system of ordinary
differential equations that approximate only mean estimates
for stochastic reaction networks.
Deterministic rate equations are generally accurate for networks
with large populations, but may incur errors when elements
are only present in low copy numbers. This thesis
presents finite state expansion, which aims to solve that problem.
It does so by converting a given reaction network into
an expanded one with additional species and reactions such
that the overall stochastic behavior is preserved. The resulting
rate equations, however, may enjoy increased accuracy.
Several tests on example models show that finite state expansion
proves competitive with other state-of-the-art methods
Novel interface discretisation methods for contact mechanics
Full text linkThis thesis’ main scope is the presentation of two different
methodologies for the analysis of contact problems involving morphologically complex or rough surfaces. Both approaches rely on the Finite Element Method (FEM) as the
chosen computational framework. They hinge on the definition of an interface finite element used to model the space encompassed by two solids incontact. This kind of interface element is shared with the field of non-linear fracture
mechanics, employed for the simulation of non-linear crack
growth according to Cohesive Zone Model (CZM). Here, for
the first time, the formulation is extensively applied to contact mechanics. With no further modifications, the interface
element is suited for the solution of contact problems involving smooth and conformal interfaces, exploiting a node-to node
approach and a penalty formulation for the enforcement
of the contact constraints. The element is enriched with specific characteristics that allow for the solution of rough contact problems yet maintaining a very simple mesh discretisation, both using a single-scale and a multiscale approach. In
the single-scale approach, a novel methodology is exploited
that considers an equivalent flat interface and accounts for
the actual geometry by a suitable correction of the standard
normal gap. In the multi-scale approach, the Boundary Element Method (BEM) is exploited for solving, at a micro-scale,
the normal contact problem of a rough rigid indenter making contact with an elastic half-space, according to a far-field displacement determined by the deformation imposed at a
macro-scale. The solution in terms of averaged pressure and
mean separation is then passed back to the macro-scale
The Bitcoin transaction networks
Full text linkThe topic of the thesis is the analysis of the most popular
cryptocurrency from a network perspective. Specifically, it
focuses on a bunch of network representations of the Bitcoin
digital transactions that are studied at different scales (micro,
meso and macro) by employing tools from physics, statistics
and economics. The thesis is divided into five chapters.
Chapter 1 introduces the broad field of cryptocurrencies. Chapters
2 and 3 are dedicated to the understanding of the network
properties of the Bitcoin cryptocurrency from both a
static and a dynamical perspective and to the investigation of
the relationships between the latter ones and purely financial
quantities like the BTC price. Chapter 4 is dedicated to the
study of the Bitcoin Lightning Network, a recently-developed
protocol to speed up the blockhain-based payment system
Bitcoin rests upon. Chapter 5 illustrates the substantial modifications
that have been required by state-of-the-art algorithms
to solve null models for networks on very large networks - as
the ones characterizing Bitcoin throughout its entire history
Arbitrage in the Bitcoin ecosystem: an investigation of the Mt. Gox exchange platform
Full text linkThe purpose of the thesis is to identify and describe the arbitrage
activity conducted in the Bitcoin ecosystem at its early
stages. This work is the first attempt in the literature to investigate
empirically the individual behavior of the arbitrageurs,
and provides evidence that they are few and sophisticated.
I exploit a dataset containing the history of trades executed
within the exchange platform Mt. Gox between 2011 and 2013.
I follow and improve upon the established methods to preprocess
the data by proposing a new approach whose validity
is documented extensively. Crucially, trades are labelled
with user specific identifiers, allowing to reconstruct the individual
sequences of actions and thus to identify arbitrageurs,
and explicit transaction costs are accounted for. The core of
the work is thus the implementation of two novel methodologies
that aim at identifying the triangular arbitrage activity
within the Mt. Gox platform and the two-point arbitrage
across Mt. Gox and two counterpart exchanges, Bitstamp and
BTC-e. In the former I focus on the mispricings of the bitcoin
price denominated in different fiat currencies; in the latter, I
compare differences in price - across Mt. Gox and the counterpart
exchanges - denominated in the same fiat currency.
I classify as arbitrageurs respectively 23 and 49 users, for a
total of 72. A comparison of aggregate statistics between arbitrageurs
and non arbitrageurs is given and discussed. This
work represents the first empirical contribution on arbitrage
at the micro level that goes beyond anecdotal evidence: the
findings challenge the textbook definition of arbitrage and
demonstrate that arbitrage is conducted by a limited number
of sophisticated and specialized investors