Presentare la propria linea di ricerca in modo da favorire collaborazioni tra ricercatori del dipartimento

Fornire informazioni utili ai ricercatori su come presentare i propri lavori per ottimizzare i risultati della prossima VQR

Maria Infusino

Titolo: A few moments with the problem of moments

Abstract: "Moments" in mathematics are quantities associated to a probability measure containing the most relevant information about it, such as the mean, the variance, the skewness. The problem of moments, also known as moment problem, is the elegant question of establishing whether there exists a probability measure having some prescribed moments. A fascinating feature of this problem is its great modeling power, which has been attracting scientists for more than a century in the most different scientific areas. This talk aims to give a taste of the immense beauty of the moment problem with a focus on the exciting connections between finite and infinite dimensional moment problems. For example, we will see how the problem of counting snakes in a pit is related to the problem of describing infinite particle systems.

Marco Livesu

Titolo: Fabrication-Aware Shape Decomposition

Abstract: Turning a digital 3D shape that lives in our computer into a real object produced with some manufacturing hardware is a complex process that requires both theoretical knowledge and practical insight. Part of this complexity depends on the constraints imposed by the fabrication technology, which severely restrict the class of shapes that can be produced with it. Some of these constraints are "hard", meaning that shapes not fulfilling them cannot be fabricated; some others are "soft", meaning that fabrication is still possible but it introduces downsides, such as increased production cost or machining time, or decreased geometric accuracy. An effective way to address both hard and soft constraints consists in splitting the 3D shape into simpler pieces, such that each sub-component is best suited for fabrication with the specific technology at hand. In this talk I will firstly present the major manufacturing paradigms. Both additive and subtractive manufacturing technologies will be considered. I will then review the most relevant approaches for shape decomposition for digital fabrication, highlighting the computational challenges and the necessary tradeoffs that are used in practical algorithms.

Francesco Beccuti

Titolo: Reflections of university students in mathematics on a simple problem involving uncertainty

Abstract: How do university students in mathematics think about a simple problem involving coin tosses? We answer to this question by analyzing and comparing the written reflections of master's students and future bachelor's students in mathematics in terms of recency and equiprobability effects (understood non-normatively). In general, we observe a major tendency of students towards equiprobability responses in connection to the display of knowledge of basic concepts of mathematical probability. The comparison between the two groups of students further shows that the influence of a university education in mathematics on the equiprobability effect is overall limited. This suggests that the students' tendency towards equiprobability responses was acquired during their previous experience in compulsory schooling.

Mirko Marras

Titolo: Ethical Personalization in Ranking Models

Abstract: Personalized ranking models play a pivotal role in shaping user experiences and influencing decision- making processes. However, the growing dependence on these models has raised pressing concerns surrounding fairness, transparency, and other ethical considerations. In this talk, we highlight the need of integrating ethical principles into the development and deployment of personalized ranking strategies. First, we delve into the challenges posed by traditional ranking approaches, which often prioritize performance metrics without adequately addressing the ethical dimensions of personalization. Then, through real-world exploratory examples and emerging algorithmic solutions, we highlight the complex interplay between fairness, explainability, and personalization. These key dimensions of ethical personalization are crucial for ensuring that ranking models do not inadvertently propagate bias, discrimination, or opacity in their results. Finally, we present some key open questions and future directions in this dynamic field, stimulating a shift towards the adoption of more responsible and transparent ranking models, where ethical considerations are at the forefront of design and implementation.

Annullato 

09 gennaio 2024

Alessandro Giuliani

Titolo: Leveraging Large Language Models for Knowledge Representation: Multidisciplinary Approaches with Emphasis on Cultural Heritage

Abstract:The main research activities concern extracting and representing information through proper knowledge bases, i.e., Knowledge Graphs (KGs), across various domains, with a particular focus on cultural heritage. In detail, the research aims to devise and test innovative methods for generating KGs by integrating large language models. Harnessing the capabilities of such advanced resources allows the enhancement of the automated synthesis of structured knowledge from structured or unstructured data sources. The research methodology involves exploiting state-of-the-art language models, such as GPT or Llama, to process and understand textual information at a nuanced level. The inferred insights are then structured into knowledge graphs, facilitating a readable and interconnected representation of information. While the research outcomes may address broader interdisciplinary applications, special attention is reserved for cultural heritage. Cultural heritage encompasses various information, from historical artifacts and artistic creations to linguistic expressions and societal practices. The findings of this research have implications for the development of advanced information systems, recommendation engines, or educational tools. Creating KGs enriched with contextualized information enhances the accessibility and interpretability of vast datasets, fostering a deeper understanding of complex subjects across multiple disciplines.

Silvia Columbu

Titolo: Latent variable models and data clustering: from standard to complex structures

Abstract: The identification of groups or clusters in data is often essential for the full understanding of relationships and structures hidden in observations. There exist many approaches to cluster detection and can be divided into two big groups: heuristic and model-based ones. The class of model-based approaches requires the assumption of a probability model whose choice is driven by the nature of the data. These methods offer the advantage of considering the theoretical framework of statistics that includes robust parameter estimation strategies, inference on the number of clusters by model selection criteria, and evaluation of uncertainty in cluster membership. The basic model is that of a finite mixture, that can be then extended and adapted to deal with more complex data structures. In this talk I will give a panoramic of standard finite mixture models and then introduce some possible extensions to address different situations: hierarchical data, simultaneous classification of different observations, clustering of nodes in a network.

Sebastian Podda

Titolo: Corporate Risk Stratification through Interpretable Deep Learning Approaches

Abstract: The seminar will be focused on presenting innovative deep learning approaches used to identify potential threats to financial sustainability for non-financial companies. As most state-of-the-art tools outcomes are often difficult to understand even for experts, this research area aim to explore an easily interpretable visualization of financial statements, projecting each company in a bi-dimensional space according to an autoencoder-based dimensionality reduction matched with a nearest-neighbor-based default density estimation. In the resulting space, the distress zones, where the default intensity is high, appear as

homogeneous clusters and can be easily identified. Companies located in a distress zone could then take the appropriate measures of intervention on liquidity, capital, and governance for adjusting their business risk trajectory

20 febbraio 2024

Cancellato

Serena Sanna

Titolo: Enhancing biomedical research with mathematics.

Abstract: Human diseases are complex and can vary significantly from one person to the next. Despite this, medical care today is based on treatments that work best on average for a large group of patients. Precision medicine aims to reshape medical care away toward a customized approach that better takes into account the individual uniqueness. To make this vision a reality, researchers are using mathematical and statistical tools to handle and efficiently analyze the large increasing amounts of medical and biological data. I will discuss my personal experience as mathematician working in the biomedical research and highlight present challenges and future perspectives.

Livio Pompianu

Blockchain for industry: analysis, opportunities and perspectives

Abstract: Blockchain is a distributed ledger technology that enables secure and transparent recording of transactions across a network of computers. No single entity controls the blockchain: it operates on a peer-to-peer network where participants verify transactions, with features including decentralization, immutability, transparency and auditability, making transactions more secure and tamper-proof. Although it became famous for cryptocurrencies, blockchain favoured the development of numerous areas such as financial and social services, healthcare, etc. However, several open challenges slow its full integration with a variety of industrial technologies (such as cloud, AI and IoT), preventing its widespread deployment. In this seminar, I will first introduce this technology and its significant paradigms. Then, I will review the different research branches involving blockchain-based solutions and their applicability in the most relevant industry applications. Finally, I will focus on the most important open questions and future direction in this field.

Silvia Frassu

Basic concepts on chemotaxis models from biomathematics

Abstract: Chemotaxis is the movement of an organism or entity in response to a chemical stimulus. The mathematical formulation idealizing this mechanism dates back to the 1970's, when Keller and Segel described it in form of two coupled Partial Differential Equations. In this seminar we will present some of these models, and we will discuss properties of related solutions.

Ludovico Boratto

La produzione scientifica come strumento di crescita professionale, del dipartimento e dell’università

Abstract. In questo seminario, contestualizzeremo la produzione scientifica di ognuno rispetto agli obiettivi di crescita professionale (Abilitazione Scientifica Nazionale, ASN) e di crescita del dipartimento e dell’università (Valutazione della Qualità della Ricerca, VQR). Proseguiremo condividendo consigli per fare in modo che il Dipartimento di Matematica e Informatica sia un ambiente in cui la crescita di tutti sia resa possibile. Concluderemo con un dibattito rispetto a impressioni, dubbi e ostacoli che si possono incontrare da questa prospettiva.
 

Simone Zanda

Optimizing Goods Distribution with Linear Optimization

Abstract: This presentation explores the use of linear optimization to tackle the complexities of goods distribution, a key aspect of supply chain management. We will discuss how linear optimization offers solutions to enhance distribution efficiency, focusing on the latest algorithms that provide both exact and heuristic strategies. The talk aims to highlight the practical impact of these optimization techniques in improving distribution systems and the role of technological advancements in overcoming logistical challenges. By emphasizing the need for ongoing research and interdisciplinary collaboration, we aim to showcase the potential for linear optimization to address current and future logistics issues.

Silvia Maria Massa

Sensor-based artificial intelligence to support people with disabilities

Abstract: A substantial portion of the world’s population deals with disability. Many disabled people do not have equal access to healthcare, education, and employment opportunities, do not receive specific disability-related services, and experience exclusion from everyday life activities. One way to face these issues is through the use of healthcare technologies. Unfortunately, there is a large amount of diverse and heterogeneous disabilities, which require ad-hoc and personalized solutions. Moreover, the design and implementation of effective and efficient technologies is a complex and expensive process involving challenging issues, including usability and acceptability. We will discuss the use of sensors combined with signal processing methods and artificial intelligence algorithms to support people with different disabilities.

Stefano Bonzio

Non-classical logics and how to study them algebraically

Abstract: 

Logic is an ancient discipline, whose founding father is traditionally recognized to be Aristotle. Naively defined as the “science of correct reasoning”, logic lies in the intersection of mathematics, philosophy and computer science.
The current research in logic breaks down in different subfields, including set theory, model theory, proof theory, the theory of computability and many more. Although logicians are interested in apparently distant topics, it is commonly accepted that a logic is a substitution invariant consequence relation over a formula algebra. Once an algebraic language is fixed, a logic (i.e. a consequence relation) can be introduced in several ways: for instance, as induced by certain calculi or by logical matrices, i.e. pairs formed by an algebra and a filter (a set of truth values). The former approach is syntactical, while the latter is semantical, where the role of algebras is precisely that of providing “meaning” to the operations in the language. In order to make the audience familiar with the difference between syntax and semantics of a logic, in the first part of the talk, I will recall some equivalent definitions of classical (propositional) logic: via syntactic calculi, such as Hilbert-style, natural deduction and sequent calculi, and via its (equivalent) algebraic semantics, the variety of Boolean algebras. Afterwards, I will try to give an idea of the field in which my research interests fall, namely non-classical logics, approached by using the tools of universal algebra and algebraic logic. I will provide a brief review of some non-classical logics, including intuitionistic, many-valued and modal logics, and show how algebraic techiniques come into play in their study.

Manuela Sanguinetti

TBA

Alessandro Buccini

From effect to cause, how to turn back time (or, at least, try to)

Abstract We are often faced with the problem of recovering an unknown signal from measurements not of the desired quantity itself, but of some its effects. This problems are referred to as inverse problems. These are often ill-posed, meaning that they are extremely sensible to the presence of perturbation in the measured data. Naive approaches for their solution usually provide poor approximation of the desired signals. As one might expect, "inverting time" is not an easy task. In this talk we discuss some approaches to tackle these problems and obtain good results. We will show examples where the function that links cause and effect is both linear and non-linear. The two cases are different, but they share several similarities and it is possible to exploit methods developed for the first one to improve solution algorithms for the second one.