PhD courses
Alberto Tesi
Date: November 25, 27
25 Nov 14,15 room S12 Santa Marta
27 Nov 9:00 room “Caminetto” Santa Marta
Hours/CFU’s: 8/2
Curriculum: Control, Optimization and Complex Systems
more details in email
Carlo Carobbi
Date: 5/12/24; 9-13 am
Hours/CFU’s: 4/1
Room: 177 Santa Marta
on line: https://meet.google.com/rmm-uaoa-vph
Curriculum: Electronics, Electromagnetics and Electrical Systems
Abstract:
Technical standards frequently provide support to, if not the basis of, academic studies and research in all the fields of engineering. Many, probably the majority, of the thesis works, PhD dissertations and scientific papers in engineering take advantage, inspiration or are even devoted to the content of specific technical standards. Nonetheless, few of the authors of such works are aware about the process leading to the writing and publication of a technical standard, its maintenance, the context in which technical standards are used by industry and referred to by regulations issued by national and regional authorities. Therefore, generally the outcomes of these studies have no impact on standards’ development. There are many reasons for this, such as academic research results are not brought to the attention of standard organizations, or do not address the actual needs of users of technical standards. The scope of this presentation is to provide information about the work of standard organizations and the context in which technical standards are applied, particularly in the electrical and information engineering sector.
Laura Carnevali
Date: 09,11,16,18 December, Santa Marta
Hours/CFU’s: 12/3
Date/Room/Time:
09/12/24, Santa Marta, aula 049, 10:15-13:15
11/12/24, Santa Marta, aula 035, 10:15-13:15
16/12/24, Santa Marta, aula 173, 09:00-12:00
18/12/24, Santa Marta, aula 173, 09:00-12:00
Curriculum: Informatics
Abstract: This course introduces the fundamentals of Markov Decision Processes (MDPs), a powerful framework for modeling of sequential decision problems under uncertainty. Topics include (but are not limited to): syntax and semantics of MDPs, adversaries, probabilistic reachability of MDPs, applications of MDPs using the PRISM tool.
Simone Magistri
Date/room:
08/01/2025, 14:00 – 17:00, Aula 49, Santa marta.
10/01/2025, 14:00 – 17:00 , Aula 49, Santa Marta.
13/01/2025, 14:00 – 17:00, Aula 175, Santa marta.
17/01/2025, 14:00 – 17:00, Aula 49, Santa marta.
Hours/CFU’s: 12/3
Curriculum: Computer Engineering
Abstract: The widespread adoption of Artificial Intelligence in real-world applications has been driven by the ability of Deep Neural Networks (DNNs) to solve increasingly complex problems in various fields, such as computer vision and natural language processing. However, the current DNN learning paradigm is static: DNNs are typically trained for single tasks with fixed datasets, making them rigid and prone to becoming outdated as real-world data and tasks evolve, particularly in sectors like transportation, healthcare, and robotics. Updating models often requires costly re-training on both old and new data, which is computationally and environmentally expensive due to the large number of parameters and the demanding training procedures involved. Additionally, privacy concerns may prevent access to old data, making re-training unfeasible. Naively updating models solely on new tasks leads to catastrophic forgetting—where models lose prior knowledge when learning new information. Continual Learning, also known as Incremental Learning, is a field of machine learning aimed at developing adaptive learning paradigms for DNNs to continuously learn new tasks while preserving previous knowledge. This course explores the fundamentals of continual learning, including continual learning scenarios such as Domain Incremental, Class Incremental, and Task Incremental, as well as regularization techniques on state-of-the-art models like Convolutional Neural Networks (CNNs) and Transformers. We will also examine recent breakthroughs, open challenges, and research directions for advancing this field.
The goal of the course is to introduce the concept of Physics Informed Deep Neural Networks (PINN), discuss its implementation from scratch in PyTorch and use advanced ad-hoc developed open-source libraries such as Nvidia-modulus for addressing real-world problems in various fields (engineering, physics, oil). We discuss recent topics such as Mixture-of-Models, Neural Operators, Physics-Informed Kolmogorov-Arnold Networks and Physics-Informed Computer Vision.
– Duration of Course: 16 hrs/4CFU (for Smart Computing, Ingegneria
Where & When:
Plesso Didattico Morgagni – viale Morgagni, 44-48, 50134 Firenze (FI)
– aula 217, Tuesday 14/01, 9:30 – 11:30
– aula 219, Wednesday 15/01, 9:30 – 12:30
– aula 219, Thursday 16/01, 9:30 – 12:30
– aula 219, Monday 20/01, 9:30 – 12:30
– aula 219, Wednesday 22/01, 9:30 – 12:30
– aula 219, Friday 24/01, 9:30 – 11:30
The course will encompass:
1. An in-depth study of differential equations and their numerical resolution using PINN.
2. An exploration of the theory and applications of Physics-Informed Neural Networks to many fields (Engineering, Physics, Medicine, Oil Industry, etc).
3. Practical sessions involving real-world applications.
Each lecture will comprise a part about the theory and a half of hands-on practice over Python Scripts and notebooks.
Either the lecture slide and the lecture code will be shared, and something is already available on the course page. For more info about the course, see
https://androbomb.github.io/teaching/
https://forms.gle/waufiF6kEwTRsveK9
2. Join the Discord channel as soon as possible.
Should you require any further information or clarification, please do not hesitate to mail me at
bombini__AT__fi.infn.it
CV: INF
Hours/CFU: 12 hours /3 CFU
Time: 9.30 : 12.30
For other details see:
Lorenzo Mucchi
Stefano Caputo
Title: Fundamentals of Physical Layer Security
Curriculum: Telecommunications and Telematics
Hours: 8 / 2 CFU
Dates: 28-29 GEN 9.30-13.30
Room: 035 Santa Marta
Abstract: Physical layer security (PLS) is a novel paradigm that aims to achieve secure wireless communications by exploiting the physical properties of the transmission channels, such as noise, interference, and fading. Unlike conventional cryptographic methods, which rely on mathematical algorithms and secret keys, PLS uses signal design and processing techniques to degrade the signal quality of the eavesdroppers and realize keyless secure transmission. PLS offers several advantages, such as avoiding the difficulties in key distribution and management, providing flexible security levels through adaptive transmission design, and leveraging the features of next-generation wireless networks, such as spatial diversity, cooperation, and cognition. This PhD course provides an overview of the main techniques of PLS and illustrates the applications of PLS in 5G and 6G networks, highlighting the challenges and opportunities of physical layer security in the contemporary landscape.
Stefano Caputo
Date: 3, 5, 7 Feb 2025
Time/Room: 9:30 – 13:30, 046 Santa Marta
CV: TLC per 2 CFU; EEE per 1 CFU
CFU: 3
Abstract: The course will give to the students the fundamentals of visible light communications (VLC). VLC refers to a data communications medium using visible light between 400 THz (780 nm) and 800 THz (375 nm). Low-cost wireless communication network can be created using VLC, often known as Li-Fi systems. LiFi is a potential solution to the shortage of global wireless radio spectrum. The light can be used as a communication medium for ubiquitous computing, because light-producing devices (such as indoor/outdoor lamps, TVs, traffic signs, commercial displays, car headlights/taillights, etc.) are used everywhere. The course will provide a general background on LiFi technology, discussing the major advantages and existing challenges to technology integration in 6G mobile Networks. Recent key advancements in physical layer techniques, such as localization, ultra-high-speed indoor connection, ultra-low latency outdoor link, will be discussed.
Luigi Chisci
Note: students are invited to register Here
Date: 10-18 / 2 /2025
Hours/CFU’s: 20/5
Date/Time:
10/2/2025 Lunedi ore 9-13
11/2/2025 Martedi ore 9-13
13/2/2025 Giovedi ore 9-13
17/2/2025 Lunedi ore 9-13
18/2/2025 Martedi ore 9-13
Curriculum: Control, Optimization and Complex Systems
Abstract: This course aims to provide both theoretical and practical tools to tackle estimation problems encountered in several areas of engineering and science. In particular, it is shown how to formulate such estimation problems as instances of a general dynamical system state estimation problem and how to derive the mathematical solution of the latter problem. Then it is shown that, for a linear Gaussian system, such a solution yields the well known Kalman filter. Further, approximate techniques (e.g. extended and unscented Kalman filters, particle filter, etc.) are presented for the case of nonlinear and/or non-Gaussian systems, for which an exact closed-form solution cannot be found. To conclude the theoretical part, theoretical limitations (i.e. the Cramer-Rao lower bound) on the quality of estimation are discussed. In the second part of the course, we illustrate some applications of linear/nonlinear Kalman filtering (e.g., tracking, robotic navigation, environmental data assimilation).
Franco Bagnoli
Date: 12/2, 19/2, 26/2, 5/3
Hours/CFU’s: 8/2
Time: 9:30 – 11:30
On-line at link: https://meet.google.com/nhe-dsor-ynb
Material: https://teaching.complexworld.net/english/cellular-automata-2025
Registration: https://forms.gle/Szby6ppW62PKcRZq5
Curriculum: Control, Optimization and Complex Systems
Abstract: Cellular automata are fully discrete systems and are used as simple models in many context, from physics to biology to computer science. They can be defined in a deterministic way, anche thus be studied as dynamical systems, extending the notions of chaos, for instance, or in a probabilistic way furnishing many examples of phase transition. The two concepts can be mixed, for instance by studying the effect of a small noise. One of the recent fields of study concerns the problem of controlling such systems, and since they are highly non-linear, standard techniques cannot be used and custom ones have to be developed.
1) Introduction to discrete modeling and simulations
2) Probabilistic cellular automata and phase transitions
3) Deterministic cellular automata, attractors
4) Control, generalizations and extensions
Pierluigi Mansueto
Date/time/Room
- 13/02/2025 – Room 046 – 14:00/17:00
- 17/02/2025 – Room 031 – 14:00/17:00
- 19/02/2025 – Room 046 – 14:00/17:00
Santa Marta
Hours/CFU’s: 8/2
Curriculum: Control, Optimization and Complex Systems
Abstract: Clustering is one of the most extensively studied problems in unsupervised learning, as highlighted by various surveys and books. Its aim is to organize a collection of elements into coherent groups called clusters: similar elements should be assigned to the same cluster while different elements should be in different clusters. This course concerns the analysis of the main models and approaches for clustering problems, with a particular focus on one of the most recognized formulation, i.e., the Minimum Sum-of-Squares Clustering (MSSC). In addition to the well-known K-MEANS, we will also examine: methodologies that do not require the number of clusters (k) a priori; novel meta-heuristics to improve the K-MEANS performance; one of the (relatively) new frontiers for this topic, that is, semi-supervised clustering.
Andrea Tani
Date:
20 Feb. 2025 9.30 – 13.30 room 171 SM
21 Feb. 2025 14.00 – 18:00 room 171 SM
CFU: 2
CV: TLC
room: 171
Abstract:
The ever-increasing demand for bandwidth in wireless communication systems, driven by multimedia and Internet of Things (IoT) applications, has led to a significant shortage of available spectrum resources as reported by the U.S. Federal Communications Commission, the underutilization of licensed spectrum due to regulated access has proven to be “a more significant problem than the physical scarcity of spectrum. This observation has highlighted the need to move away from static resource allocation strategies and has driven intense research into dynamic spectrum sharing among systems with heterogeneous radio access technologies and different priorities for access to licensed and unlicensed bands. These research efforts have culminated in the development of cognitive radio (CR) technology, which seeks to exploit underutilized spectral resources in the frequency, time, and space domains by reusing them opportunistically. Notably, dynamic spectrum sharing and management have been recognized as potential enablers for 6G, falling under the category of “new spectrum.” At the core of cognitive radio lies spectrum sensing, tasked with determining whether a particular portion of the spectrum is “available” or not. In other words, the goal is to discriminate between two mutually exclusive hypotheses. Following a brief introduction to the application scenarios of the CR paradigm, the first part of the course provides the theoretical foundation of spectrum sensing within the framework of detection theory, with a particular focus on blind techniques. The second part of the course addresses the challenges associated with applying blind spectrum sensing techniques in the context of in-band full-duplex technology, as well as their operation at millimeter-wave and terahertz frequencies, both of which are envisioned as potential 6G enablers. Finally, we present the application of spectrum sensing in the integration of unmanned aerial vehicles UAVs into 5G/6G networks, addressing challenges such as dynamic spectrum access in multi-UAV scenarios
Carlo Odoardi, Lorenzo Capineri
Date: 9,23 Apr 2025
Room/Time:
– 9/4 10.15-13.15, room 48 (or sala saminetto), Santa Marta
– 23/4 14.30-17.30, room 48 (or sala saminetto), Santa Marta
Hours/CFU’s: 6/1
Curriculum: Soft Skills
Abstract:
Part 1. Strategic Human Resource Management for Innovation Prof. Carlo Odoardi
The strategic management of the human capital present in an organization falls within the framework defined as “organizational innovation” which can be defined as the connection of people to organizational objectives and goals in order to improve performance and develop a culture aimed at enhancing skills and professionalism to support innovation, flexibility and competitive advantage in continuous collective or organizational learning. An organization focused on the enhancement of people must necessarily equip itself with a series of new systems and measurement methods to verify and monitor the innovative variables necessary to determine specific professional behaviours at an individual, team and organizational or managerial level in an integrated vision. Today, research and experimentation carried out over the last twenty years in work contexts has highlighted the strategic importance of the interaction and integration between “organizational innovation and technological innovation”. The field of intervention research has produced a series of models, methods and tools for the development of organizations in an innovative way and above all to outline new organizational models (agile organizations or smart organizations) and management models (new managerial and leadership models for innovation) where the system of professional relational networks are central to dealing with continuous metamorphoses.
Part 2. The complexity of the role of the future engineer Prof. Lorenzo Capineri
This course focuses on the complexity of STEM education, especially for engineering courses, considering the present and future role of these professionals in society. The high specialization of technologies has required vertical education in engineering topics, while the role of engineers of the future will be broader to solve societal problems at a global level. The ability to fill this role can be improved by stimulating the education of engineers with humanistic and social topics, multicultural teamwork and preparation to stimulate creativity and innovation within a laboratory community. In this path we find the relevance of Adriano Olivetti training process. This course reports some experiences and methodologies of innovative teaching that can be useful for the design of university teaching courses in STEM subjects”. Contributions from large industry managers is foreseen.
Ultrasound medical imaging: an engineer’s perspective
Proff. Ramalli/Meacci
- 6 May 08:30 – 12:30
- 9 May 13:30 – 17:30
Room: 033 Santa Marta
CV: Electronics, Electromagnetics and Electrical Systems
Hours/CFU: 8/2
abstract:
Ultrasound imaging techniques are a widely used diagnostic tool in medicine. They owe their success to a series of features that make them ideal for medical applications. Indeed, they use a form of energy that does not entail harmful effects on biological tissues. Moreover, these techniques can be implemented in relatively low-cost and low-size systems working in real-time, which can be useful to perform exams directly at the bedside or in the operating room.
The course will provide the basics of the physics of ultrasound, starting from the ultrasound wave generation to the concepts of reflection and backscattering. Similarities between ultrasound echograpic and radar systems will be highlighted.
Then, the course will focus on the signal and image processing techniques used for morphological and motion (tissue and blood-flow) imaging. Technical requirements, limitations, and solutions, which impact the final image quality, will be discussed.
Finally, the course will give an overview of advanced, state-of-the-art equipment and techniques for biomedical applications, including ultrafast imaging systems and 3-D imaging.
The course includes a demo session with an ultrasound scanner fully developed by the Department of Information Engineering of the University of Florence.
Franco Bagnoli
Date: 7/5, 14/5, 21/5, 28/5, 4/6, 11/6
Hours/CFU’s: 12/2
Time: 14:30-16:30
on line at link: https://meet.google.com/eox-upfq-pxf
Material: https://teaching.complexworld.net/english/from-babbage-to-chatgpt-2025
Registration: https://forms.gle/M7Bd1RCh6vZQMZeU9
Curriculum: Soft Skills
Abstract: We shall review the evolution of computers and their applications, from the first mainframes dedicated to computing, to the evolution in the business world, the birth of the Internet, the switch to personal computers, and finally to the internet & microprocessor world of today. In parallel, we shall examine the development of operating systems and computer languages, the social impact and its driving force, the connections with literature and science fiction.
Complex-valued neural network: theoretical aspects and applications for failure prevention in electrical systems.
Marco Bindi
Room / Time TBD
CV: EEE
CFU:1
DATE: 30 May 09:00-13:00
Room: Sala Caminetto Santa Marta
This course is structured into two main parts and will be held in a single day, lasting four hours. The first part provides a theoretical introduction to Multi-Layer Neural Networks with Multi-Valued Neurons (MLMVN), highlighting their key properties and differences from conventional real-valued feedforward neural networks. The second part focuses on practical applications of MLMVN in classification problems related to fault diagnosis in medium-voltage electrical cables, DC/DC power converters, and Power Quality assessment. Additionally, live demonstrations of the training process using a MATLAB application may be presented during the session.
Electrical load forecasting for applications in the Smart Grid
M. Intravia
CV: EEE
CFU:1
DATE: 26 May 09:00-13:00
Room: Sala Caminetto Santa Marta
The course begins by highlighting the importance of electrical load forecasting in Smart Grids, particularly in relation to energy management systems. Forecasting plays a crucial role: by predicting energy demand in advance, utilities and operators can enhance efficiency, integrate renewable energy sources more effectively, and prevent overloads or shortages.
The course then consists of a first theoretical part, where the fundamental concepts of forecasting are introduced, along with the most commonly used techniques in this field. In particular, both classical (or statistical) approaches and more advanced approaches based on neural networks, such as LSTM networks, will be considered.
In the second part, practical examples of these algorithms applied to electrical load forecasting will be presented. Additionally, the implementation of these algorithms in Python code will be demonstrated.
UAV Communications
Andrea Tani
11-12 June 9.30 am
10/06 9:30-13:30 Room 171 Santa Marta
12/06 9:30-13;30 Room 173 Santa Marta
Hours/CFU: 8/2
CV: TLC
abstract:
Nowadays, Unmanned Aerial Vehicles (UAVs), also commonly known as drones have discovered a wide range of applications in various domains, including aerial inspections, photography, precision agriculture, traffic management, search and rescue operations, package delivery, and telecommunications, among others.To achieve ubiquitous and high-data-rate services in challenging scenarios, new-generation wireless networks must embrace innovative technologies. The integration of terrestrial and aerial nodes to create a vertical heterogeneous network offers flexible and reliable mobile communication infrastructure. Deploying UAVs as aerial base stations in wireless communication systems promises cost-effective connectivity, particularly in areas without existing infrastructure coverage. Low-altitude UAVs outperform terrestrial communications in terms of deployment speed, flexibility, and quality of service, thanks to short-range line-of-sight links.The first part of the course focuses on examining empirical models that characterize both Air-to- Air (AA) and Air-to-Ground (AG) propagation channels, considering various scenarios (Rural, Sub- Urban, Urban). It also provides an overview of UAV-aided wireless communications, addressing typical use cases like UAV-aided ubiquitous coverage and UAV-aided relaying. Special emphasis is placed on addressing the challenges related to trajectory planning and the deployment of UAV networks to achieve optimal wireless coverage.The second part of the course deals with the UAV-to-UAV communications, UAV-Cellular spectrum sharing, and provides insights into techniques for identifying and countering malicious jamming attacks.
Architectures and Protocols Design Towards Quantum Internet
Roberto Picchi
room: 171 Santa Marta
Time: 9-13
Date 25-26 June
CV: TLC CFU: 2
Abstract: Quantum mechanics revolutionized the understanding of the physical world, yet the realization of quantum computing remained theoretical for decades. Richard Feynman introduced the concept of quantum computing, demonstrating its advantages over classical Turing machines, which was later formalized by David Deutsch in 1985. Subsequent research further refined quantum computing models, highlighting their fundamental differences from classical computation. The rapid advancement of telecommunications has driven increasing interest in quantum communication networks. While terrestrial quantum communication via Optical Fibers (OFs) suffers from significant signal loss, necessitating costly repeaters, Quantum Satellite Networks (QSNs) present a promising alternative. These networks facilitate long-distance entanglement distribution and enable intercontinental quantum communication. Research on Low Earth Orbit (LEO) quantum satellite backbones aims to interconnect quantum servers, enhancing computational capabilities. A critical challenge is optimizing End-to-End (E2E) entanglement generation. By leveraging Software-Defined Networking (SDN), proposed architectures minimize the number of hops while maximizing network capacity, striking a balance between performance and cost across centralized and distributed models in various satellite constellations. Studies in the literature highlight the benefits of embedding a Control Plane (CP) within the satellite constellation. A two-tier CP model has been proposed, with a Master Control Station (MCS) on the ground managing the network, while satellite-integrated CPs handle entanglement generation. Additionally, protocols for E2E entanglement generation have been introduced and evaluated. Furthermore, research on drone-based Metropolitan Quantum Networks suggests they offer a flexible, cost-effective solution. SDN has been shown to play a crucial role in managing Quantum Drone Networks (QDNs), enabling efficient entangled pair distribution. An SDN-based architecture for Metropolitan Quantum Drone Networks (MQDNs) has been proposed, incorporating an entanglement generation protocol and an optimization framework. Performance evaluations in literature assess fidelity, entanglement rate, and overhead, demonstrating the feasibility of QDNs for distributed quantum computing and Quantum Key Distribution (QKD).
Angeli
CV: Control, Optimization and Complex Systems
Time: 23 july 2025, time 10:00 – 12:00 and 14:00 – 16:00
Room 048 Santa Marta
Abstract:
Chemical reaction networks are a modelling tool for the complex molecular interactions underpinning life at the cellular level. They bear close resemblance to Petri Nets and are characterized, at the dynamical level, by the presence of strong nonlinearity.
As such they have attracted the attention of both biologists and engineers, trying to understand how their structure can affect their functionality and their dynamical behaviors.
The course will introduce the basic models of chemical reaction networks and illustrate some robust tools for the study of their dynamics using ordinary differential equations.
*: PhD course proposal, still to be approved by the PhD committee
PhD Seminars
Normally 1 CFU
Sala Caminetto Santa Marta,
17/4/2025 ore 11
Prof Joao Magalhães (Universidade NOVA de Lisboa)
11 April – 11:30 – Room 008 Centro Didattico Morgagni, V.le Morgagni 40-44
Title
Contrastive Beam Diffusion Models for Decoding Visual Sequences
Abstract
While diffusion models excel at generating high-quality images from text prompts, they struggle with visual consistency in image sequences. Existing methods generate each image independently, leading to disjointed narratives – a challenge further exacerbated in non-linear storytelling, where scenes must connect beyond adjacent frames. We introduce a novel beam search strategy for latent space exploration, enabling conditional generation of full image sequences with beam search decoding. Unlike prior approaches that use fixed latent priors, our method dynamically searches for an optimal sequence of latent representations, ensuring coherent visual transitions. To address beam search’s quadratic complexity, we integrate a contrastive mechanism that efficiently scores search paths and enables pruning, prioritizing alignment with both textual prompts and visual context. Human evaluations confirm that our approach outperforms baseline methods, producing full sequences with superior coherence, visual continuity, and textual alignment. By bridging advances in search optimization and latent space refinement, this work sets a new standard for structured image sequence generation.
Bio
João Magalhães is a Full Professor at the Department of Computer Science, Universidade NOVA de Lisboa, is national co-Director of the CMU-Portugal partnership and leads the Multimodal Systems Group at NOVA LINCS. He holds a PhD from Imperial College London (2008) and conducts research at the intersection of AI, vision, and language, focusing on generative models, controllable LLMs, multimedia search, multimodal conversational AI, and temporal models. João has coordinated and contributed to numerous international projects with partners like BBC, Amazon and Google. He has held key organizational roles in top-tier conferences, including General Chair of ACM Multimedia 2022 and PC Chair of ACM Multimedia 2026. His work has earned multiple awards, including first and second place in the Amazon Alexa Taskbot Challenge. He also contributed to MPEG-7 and MPEG-21 standards during his time in industry. João is currently a member of the ACM Multimedia Steering Committee.
Time 14.30
Room 175 Santa Marta
2 seminars:
Martina Cerulli
Approaches to Solving Convex Semi-Infinite Programs with a Non-Convex Inner Problem
Semi-Infinite Programs (SIPs) are optimization problems characterized by a finite number of decision variables but an infinite number of constraints, typically parameterized over a continuous domain. These problems arise in various applications, including robust optimization, control theory, and game theory, where constraints must be satisfied over an entire range of parameters. Due to their complexity, solving SIPs requires specialized techniques that efficiently handle the infinite constraint set. We discuss a solution approach for SIPs based on the dualization of the inner problem, i.e., the problem of finding the constraint that is the most violated by a given point. After a brief introduction to SIPs and the classical solution techniques for these optimization programs, we will will present the results of the paper “Convergent algorithms for a class of convex semi-infinite programs” by M. Cerulli, A. Oustry, C. D’Ambrosio, and L. Liberti (SIAM Journal on Optimization, 2022). This paper focuses on convex SIPs with an infinite number of quadratically parametrized constraints, not necessarily convex with respect to the parameter. A new convergent approach to solve these SIPs is proposed, leveraging the dualization technique. Based on the Lagrangian dual of the lower-level problem, a convex and tractable restriction of the considered SIP is derived. We state sufficient conditions for the optimality of this restriction. If these conditions are not met, the restriction is enlarged through an Inner-Outer Approximation Algorithm, and its value converges to the value of the original semi-infinite problem. This new algorithmic approach is compared with the classical Cutting Plane algorithm on two applications: constrained quadratic regression and a zero-sum game with cubic payoff. To conclude the talk, we will give some hints on how to deal with the main challenge in addressing these problems: solving the separation problem, namely, finding the most violated constrain. In “Convex semi-infinite programming algorithms with inexact separation oracles” by A. Oustry and M. Cerulli (Optimization Letters, 2024) we propose to tackle this difficulty by solving the separation problem approximately, using an inexact oracle.
———————————————
Mahsa Yousefi
Fully stochastic trust-region methods with Barzilai-Borwein steplengths
We discuss stochastic gradient methods using stochastic adaptations of Barzilai-Borwein (BB) steplengths for finite-sum minimization problems. Our approach builds on the Trust-Region-ish (TRish) framework, a first-order stochastic trust-region method based on careful step normalization. Our framework, TRishBB, is designed to enhance the performance of TRish while reducing the computational cost of its second-order variant. In this talk, we introduce TRishBB in three variants, each leveraging BB steplengths in a stochastic setting. We will highlight the theoretical foundations of TRishBB, key insights, and properties from the convergence analysis, and discuss its practical impact on machine learning applications with numerical results.
Limited memory gradient methods for unconstrained optimization
Dr. Giulia Ferrandi e il Dr. Michiel Hochstenbach
Room 327 Morgagni
Time:11:30
Abstract : The limited memory steepest descent method proposed by Fletcher (LMSD, [1]) for unconstrained optimization problems stores a few past gradients in a matrix, to compute multiple stepsizes at once. In the quadratic case, the matrix of gradients forms a basis for a Krylov subspace, and the stepsizes are the reciprocals of a few Ritz values of the Hessian matrix. LMSD for general nonlinear functions is derived from the quadratic case, but needs some adaptations since the Hessian matrix is no longer constant. We review LMSD as presented by Fletcher, and propose some new variants. For strictly convex quadratic objective functions, we study the numerical behavior of different techniques to decompose the matrix of gradients and compute new stepsizes. In particular, we introduce a method to improve the use of harmonic Ritz values. We also show the existence of a secant condition associated with LMSD, where the approximating Hessian is projected onto a low-dimensional space. In the general nonlinear case, we propose two new alternatives to Fletcher’s method: first, the addition of symmetry constraints to the secant condition valid for the quadratic case; second, a perturbation of the last differences between consecutive gradients, to satisfy multiple secant equations simultaneously. We show that Fletcher’s method can also be interpreted from this viewpoint.
Adaptive Randomized Pivoting for low-rank approximation
room 119 Morgagni
Time 11:30
Alice Cortinovis
Abstract : We consider the problem of finding a low-rank approximation of a given matrix in an efficient way. We focus on approximations that are built from rows and columns of the matrix, starting with the column subset selection problem. We propose a randomized strategy to select suitable rows and columns, called Adaptive Randomized Pivoting. The algorithm is simple and it guarantees, in expectation, an approximation error that matches the optimal existence result in the Frobenius norm. To show the versatility of Adaptive Randomized Pivoting, we apply it to select indices in the Discrete Empirical Interpolation Method, in cross approximation of general matrices, and in the Nyström approximation of symmetric positive semidefinite matrices. In all these cases, the resulting randomized algorithms enjoy bounds on the expected error that match – or improve – the best known deterministic results.
14 Marzo 2025, ore 11:00 – 12:30
Scuola di Ingegneria, Università di Firenze
Aula Caminetto – Salone di Villa Cristina
PROF. CRISTIANO TOMASSONI
UNIVERSITÀ DI PERUGIA
IEEE DISTINGUISHED LECTURE
ABSTRACT: The Additive Manufacturing (AM) technology, also known as 3D-printing technology, offers several interesting and attractive features, including fast prototyping, geometry flexibility, easily customizable products, and low cost (in some cases). However, using such technologies for microwave devices is not straightforward as AM has not been specifically developed for microwave components, and in most cases, some adaptation and post-processing is necessary. Furthermore, there are many AM technologies available, and it is important to understand their characteristics before selecting one.
In the presentation, an overview of the different AM technologies available will be provided. Additionally, an analysis of some of the most common AM technologies used for the manufacturing of microwave components will be conducted in more detail, with the help of several examples. Several microwave components manufactured with some of the most popular AM technologies will be shown, along with a detailed description of the manufacturing process, post-processing, and all actions necessary to make the component perform well. Furthermore, it will be shown how the flexibility of this technology allows the development of new classes of components with non-conventional geometries that can be exploited to obtain high-performing components in terms of compactness, weight, losses, etc.12
Register at: https://events.vtools.ieee.org/m/469088
Mercoledì 12 Marzo ore 10, Aula Caminetto – Santa Marta
Prof. Jerzy Sawicki, Ph.D.
Center for Rotating Machinery Dynamics and Control
Cleveland State University
Control Driven Advances in Smart Rotating Machinery
Over the past three decades, significant advancements have been made in the design of rotating machinery equipped with smart components and embedded functions. These advancements have been driven by developments in actuators, sensors, and power electronics technologies, along with improvements in data acquisition, signal processing, and control theory. This presentation will provide an overview of the current state-of-the-art and showcase examples of smart technologies applied to rotating machines. Several technologies, either most recently developed or under development will be presented, all of which involve active control and techniques that are relevant to smart solutions applied to rotating machinery. Key developments will be described, and case studies from the speaker’s research will be highlighted.
The presentation will commence with an introduction to Cleveland State University and the Washkewicz College of Engineering, followed by an overview of the research activities at the Center for Rotating Machinery Dynamics and Control.
Bio
Jerzy Sawicki earned his Ph.D. in Mechanical Engineering from Case Western Reserve University, USA. He also holds a B.S. and M.S. in Mechanical Engineering and Applied Mathematics from Gdansk University of Technology and the University of Gdansk, Poland, respectively. Currently, he serves as the Department Chair and holds the D.E. Bently and A. Muszynska Endowed Chair and Professorship in Mechanical Engineering at Cleveland State University (CSU). He founded and directs the Center for Rotating Machinery Dynamics and Control (RoMaDyC) at CSU’s Washkewicz College of Engineering. From 2010 to 2020, he served as the Vice President for Research at CSU. Since 2017, he has been the Editor-in-Chief of the ASME Journal of Engineering for Gas Turbines and Power.
Dr. Jannis Kurtz Amsterdam University
Date 6 Feb. 2025
Time 14:30
room 175 Santa Marta
Abstract: In recent years, there has been a rising demand for transparent and explainable machine learning (ML) models. A large stream of works focuses on algorithmic methods to derive so called counterfactual explanations (CE). Although significant progress has been made in generating CEs for ML models, this topic has received minimal attention in the Operations Research (OR) community. However, algorithmic decisions in OR are made by complex algorithms which cannot be considered to be explainable or transparent. In this work we argue that there exist many OR applications where counterfactual explanations are needed and useful. In the first part of the talk, we translate the concept of CEs into the world of linear optimization problems and define three different classes of CEs: strong, weak and relative counterfactual explanations. For all three types we derive problem formulations and analyze the structure. We show that the weak and strong CE formulations have some undesirable properties while relative CEs can be derived by solving a tractable convex optimization problem. We test all concepts on a real-world diet problem and we show that relative CEs can be calculated efficiently on NETLIB instances. In the second part of the talk, we analyze CEs for binary knapsack problems, making a first step towards general integer problems. For special cases of the problem, we present solution methods which perform well in preliminary computational experiments.
Soft Skill
CFU 1
Profili dei ricercatori e Valutazione della ricerca
Partecipate al nostro webinar informativo su “Profili dei Ricercatori e Valutazione della Ricerca,” dove esploreremo gli strumenti e le strategie essenziali per costruire e migliorare i profili dei ricercatori. Questa sessione mira a fornire ai partecipanti le conoscenze necessarie per dimostrare e descrivere efficacemente l’impatto del lavoro di un autore.
Obiettivi Formativi:
-Costruire il proprio profilo ricercatore
-Dimostrare e descrivere i dati dell’impatto di un autore
Questo webinar è progettato per ricercatori, accademici e professionisti interessati a comprendere come sfruttare Web Of Science e InCites per la valutazione della ricerca e la costruzione del profilo.
Non perdere questa opportunità per migliorare le competenze e conoscenze nella valutazione dell’impatto della ricerca.
Data / Ora: 16 Gennaio 2025 11:00-12:00
Link di registrazione
https://clarivate.com/academia-government/events/profili-dei-ricercatori-e-valutazione-della-ricerca/
13 Dicembre 2024, ore 9:00 – 16:30
Scuola di Ingegneria, Università di Firenze
Aula Caminetto – Salone di Villa Cristina
La Giornata affronta la continuità tra didattica, ricerca e trasferimento tecnologico, incardinati nella Scuola di Ingegneria e nel Dipartimento di Ingegneria dell’Informazione, con l’industria elettronica dell’area fiorentina. Dai contributi industriali emergerà la figura professionale dei futuri ingegneri elettronici. Alla giornata parteciperanno le seguenti aziende:
Leonardo SpA; Eldes Srl; Microtest SpA; Saitec Srl; Powersoft SpA; Promel Srl; Pasquali Microwave Systems Srl; Advance Microwave Engineering Srl; Ampatech Srl; Pramatech Srl; SSE Srl; ABB e-mobility SpA.
E’ necessaria l’iscrizione: https://forms.gle/zL74NT9SEPEpJznq8
In chiusura verranno consegnati i premi di laurea attribuiti agli studenti della laurea magistrale in Ingegneria dei Sistemi Elettronici.
Hervè A. Corti, Introduction to Quantum Computatione Title
Room 051 Santa Marta
Time 10.15
Date: 11/12/24
Lavinia Amorosi
Title: A Mathematical Programming Approach to Hierarchical Clustering
date/place: December 3rd 2024, 11:30 a.m., Room 049 Santa Marta
abstract: Hierarchical clustering is a statistical technique to study the occurring groups (clusters) within a dataset creating a hierarchy of clusters. This is represented by a rooted tree (dendrogram) whose leaves correspond to the data points, and each internal node represents the cluster containing its descendant leaves. Among methods to perform hierarchical clustering, the agglomerative ones are based on greedy procedures that return a sequence of nested partitions where each level up joins two clusters of the lower partition relying on a local criterion.
In this talk, motivated by the lack of exact approaches that guarantee global optimality, we focus on a unified mathematical programming formalization for single and complete linkage procedures. We evaluate, according to different measures commonly used in this context, the dendrograms obtained from the exact resolution of the formulations and those produced by the greedy approach.
Furthermore, by exploiting the mathematical formulation, we also present a scalable matheuristic algorithm capable of generating better quality dendrograms than those produced by the greedy approach even for large size datasets.
Date/time: 26 Nov. 16:00
Room: 035 Santa Marta
Title: Image matching: vecchie glorie e nuovi orizzonti
Abstract. L’image matching (calcolo di corrispondenze) è alla base della maggior parte dei metodi di ricostruzione 3D e di stima di posa in computer vision, con applicazioni notevoli quali SfM (Structure from Motion), Visual SLAM (Simultaneous Localization and Mapping) e più recentemente NeRF (Neural Radiance Fields) and GS (Gaussian Splatting). Dopo aver introdotto le fasi principali del processo di image matching, verrà presentata l’evoluzione di tali approcci fino allo stato dell’arte attuale, mostrando i pregi e i difetti delle tecniche più rilevanti (alcune della quali basate su deep learning) e le criticità connesse ai metodi di valutazione e ai dataset impiegati. Infine verranno discusse semplici strategie generali basate sull’utilizzo di piani virtuali per il miglioramento di metodi di image matching.
E. Planas
2024, 13th Dic, 9:00 a.m., On line at link:
https://us02web.zoom.us/j/87052041160?pwd=WvmfWSNWSI48NWGhXXjfaboCZ88GOB.1
Meeting ID: 870 5204 1160
Passcode: 0uARAZ
Abstract:
This seminar introduces the National Institute of Informatics (NII) andopportunities for international cooperation with NII. NII is the Inter-University Research Institute (ROIS) for advanced ComputerScience Research and Global Science Data Infrastructure support, situated inthe heart of Tokyo, Japan. NII is ranked top 2 for publications in ComputerScience in Japan. Research activity is divided in four Departments:
1.Principles of Informatics;
2.Information Systems Architecture Science;
3.Digital Content and Media Sciences;
4.Information and Society.
The Global Liaison Offi ce (GLO) at NII has established partnerships with more than 100 MOU partners. GLO supports international exchanges via the“MOU Grant Program” under which a NII researcher can invite a colleaguefrom overseas for a stay, or send some member of his team for a researchvisit (up to 2 months, around 40 researches a year supported). GLO alsooff ers the “NII International Internship Program” in which we welcome andfund MOU partners Master’s and PhD. students for a research stay in one ofour teams (3 to 6 months, around 140 students per year).
A smart city represents an improvement of today’s cities both functionally and structurally, that strategically utilizes many smart factors, such as information and communications technology (ICT), to increase the city’s sustainable growth and strengthen city functions, while ensuring citizens’ quality of life and health. Cities can be viewed as a microcosm of “objects” with which citizens interact daily: street furniture, public buildings, transportation, monuments, public lighting and much more. Moreover, a continuous monitoring of a city’s status occurs through sensors and processors applied within the real-world infrastructure. Industrial sites represent similar scenarios, where data collected from distributed objects
allow to actuate powerful control strategies.
The Internet of Things (IoT) concept imagines all these objects being “smart”, connected to the Internet, and able to communicate with each other and with the external environment, interacting and sharing data and information. Each object in the IoT can be both the collector and distributor of information regarding
mobility, energy consumption, air pollution, as well as potentially offering cultural and tourist information.
As a consequence, cyber and real worlds are strongly linked in a smart city, such as in industrial site. New services can be deployed when needed and evaluation mechanisms will be set up to assess the health and success of the system under control. This talk will present some innovative developments in areas related to smart cities and smart industries, leveraging on the features supported by network intelligence at the edge of the network.
Soft Skills
Courses and seminars for Soft Skills organised out from Department of Information Engineering
1 CFU per 6 hours of course
BiblioTech, Bibliographic Search for Engineering PhD Students
Dispensed from Technology area Library
Courses list from “Student Learning” Florence University
Dispensed from Florence University
Courses list for Ph.D. students from Florence University
Dispensed from Florence University