Gen 142025
 

MAIN PhD Seminars 2024

Date Speaker(s)
January, 15th Sandro Gabriele Tiddia
January, 29th Andrea Azzarelli
February, 12nd Valentino Artizzu
February, 26th Simone Pusceddu
March, 12nd Nicola Piras

All the seminars at 12:30 PM in Aula II.

 

Sandro Gabriele Tiddia: LLM Agents: Definitions and Real-World Applications

In this seminar, we explore the concept of ‘agents’ in artificial intelligence (AI), with a particular focus on the role of Large Language Models (LLMs) in powering these systems. The seminar begins by discussing a real-world application where LLMs are used to build a question-answering (QA) system, showing how LLMs can function as ‘agents’ within such a system. We then examine various definitions of ‘agent’ across AI subfields and consider how agents interact with their environment, make decisions, and pursue goals autonomously. Additionally, we revisit earlier works on agency in AI, reflecting on their original, more profound ideas, and connect them to recent developments and applications of LLM-powered agents. The seminar concludes by exploring experiments and use cases from recent literature, highlighting the capabilities and potential of LLM agents across different domains. The goal is to provide a clear introduction to the concept of LLM-powered agents, their role in AI systems, and how this concept has evolved from theoretical foundations to practical applications.

 

Gen 102025
 

Interpretable and Explainable Machine Learning Models

Dr. Claudio Pomo
Politecnico di Bari

Abstract

The course focuses on methods for interpreting and explaining machine learning (ML) models, including inherently interpretable approaches and post-hoc explanation techniques. Key concepts of interpretability will be introduced, alongside the analysis of interpretable models and the application of explanation methods for complex models. The course critically evaluates existing techniques in terms of fidelity, stability, fairness, and practical utility, while addressing open challenges and future perspectives.

Schedule

The course will have a total duration of 10 hours, divided into 4 sessions. Dates and times will be announced later.

Exam

The final exam consists of a project analyzing a case study using the techniques and tools acquired during the course. The course will be held in person. Please contact me if you are interested in joining.

References

  1. Lundberg, S. M., and Lee, S.-I., A unified approach to interpreting model predictions, Advances in Neural Information Processing Systems, 2017.
  2. Ribeiro, M. T., Singh, S., and Guestrin, C., Why should I trust you? Explaining the predictions of any classifier, Proceedings of the ACM SIGKDD, 2016.
  3. Molnar, C., Interpretable Machine Learning: A Guide for Making Black Box Models Explainable, 2nd edition, 2022.
  4. Doshi-Velez, F., and Kim, B., Towards a rigorous science of interpretable machine learning, arXiv preprint, 2017.
  5. Agarwal, C., Krishna, S., Saxena, E., Pawelczyk, M., Johnson, N., Puri, I., … & Lakkaraju, H., Openxai: Towards a transparent evaluation of model explanations, Advances in Neural Information Processing Systems, 2022
 Scritto da in 10 Gennaio 2025  Senza categoria  Commenti disabilitati su PhD Course: Interpretable and Explainable Machine Learning Models
Ott 242024
 

Introduction to algebraic logic

Dr. Nicolò Zamperlin
Università degli Studi di Cagliari

Abstract

The course is an introduction to the theory of algebraizability of Blok and Pigozzi. Through an analytic study of the first chapters of Font’s handbook on abstract algebraic logic we will first introduce the elementary notions of universal algebra needed for linking together logic and algebra (closure operators and their lattices, varieties, quasivarieties and equational consequences), then building upon these notions we will consider the case of implicative logics and their algebraic properties, introducing the technique of completeness through the Lindenbaum-Tarski process. Finally we generalize these notions to the class of algebraizable logics (with a glimpse to the larger Leibniz heirarchy), with the ultimate goal of proving the isomorphism theorem and the transfer for the deduction theorem.

Schedule

The course will have a duration of 20 hours, scheduled as follows:ˆ

  • November 7, aula B, h. 15-17
  • November 14, aula A, h. 15-17
  • November 22, aula II, h. 9:30-11:30
  • November 29, aula II, h. 9:30-11:30
  • December 2, aula II, h. 9:30-11
  • December 6, aula II, h. 9:30-11:30
  • December 12, aula B, h. 15-17
  • January 15, aula B, h. 10-12
  • January 20, aula B, h. 15-17
  • February 3, aula B, h. 10-12

Exam

The final exam consists in a seminar presentation. The course will be held in person. Please contact me if you are interested in joining the course

References

  1. Bergman, C., Universal Algebra: Fundamentals and Selected Topics, Chapman & Hall Pure and Applied Mathematics, Chapman and Hall/CRC, 2011.
  2. Blok, W., and Pigozzi, D., Algebraizable logics, vol. 396 of Memoirs of the American Mathematical Society, A.M.S., 1989.
  3. Burris, S., and Sankappanavar, H.P., A course in Universal Algebra, freely available online: https://www.math.uwaterloo.ca/snburris/htdocs/ualg.html, 2012 update.
  4. Czelakowski, J., Protoalgebraic logics, vol. 10 of Trends in Logic: Studia Logica Library, Kluwer Academic Publishers, Dordrecht, 2001.
  5. Font, J.M., Abstract Algebraic Logic: An Introductory Textbook, College Publications, 2016
 Scritto da in 24 Ottobre 2024  Senza categoria  Commenti disabilitati su PhD Course: Introduction to algebraic logic
Set 232024
 

Conic Programming: Theory and applications

Prof. Benedetto Manca
Università degli Studi di Cagliari

Abstract

The course covers the theory of conic programming, starting from the simplest case of linear programming and introducing conic quadratic and semi-definite programming. The first part of the course will introduce the theoretical backgrounds needed to define the concept of conic programming. In the second part the case of conic quadratic and semi-definite programming will be addressed together with some applications.

Outline

  • From Linear to Conic Programming
  • Conic Quadratic Programming
  • The quadratic formulation of the Distance Geometry Problem
  • Semi-definite Programming
  • The semi-definite relaxation of the Distance Geometry Problem
  • Diagonally dominant matrices and positive semi-definite matrices
  • The ellipsoidal separation problem

Schedule

The course consists in 10 hours, two lectures per week. Details will be specified on the occasion of the first lecture, which will be given on October 3, 2024 at 2:30 p.m. in room B of the Department of Mathematics and Computer Science.

Exam

The final exam consists in a presentation on a specific application of conic programming (conic quadratic or semi-definite).

References

  1. Ben-Tal, Aharon, and Arkadi Nemirovski. Lectures on modern convex optimization: analysis, algorithms, and engineering applications. Society for industrial and applied mathematics, 2001.
  2. Liberti, Leo. “Distance geometry and data science.” Top 28.2 (2020): 271-339
  3. Astorino, Annabella, et al. “Ellipsoidal classification via semidefinite programming.” Operations Research Letters 51.2 (2023): 197-203.
Set 122024
 

Introduction to Kähler Geometry

Prof. Roberto Mossa, Prof. Giovanni Placini
Università degli Studi di Cagliari

Abstract

This introductory course covers some of the fundamental concepts of Kähler geometry, with particular attention to almost complex and complex manifolds, the properties of Hermitian metrics, and Kähler metrics. Starting from the basics of differential geometry, we will explore the structure of almost complex and complex manifolds. Subsequently, we will delve into the properties of Hermitian metrics, focusing on the definition and characteristics that define Kähler metrics, which play a key role in integrating the complex structure with the Riemannian one. Through concrete examples and applications, students will gain a deep understanding of these concepts, preparing them for advanced studies in Kähler geometry.

Schedule

January-February 2025 (TBA)

Exam

The final exam consists in a seminar.

 Scritto da in 12 Settembre 2024  Senza categoria  Commenti disabilitati su PhD Course: Introduction to Kähler Geometry
Ago 212024
 

Numerical Analysis with Deep Neural Networks

Prof. Yuesheng Xu
Old Dominion University and Syracuse University, USA

Abstract

This four-talk lecture sequence aims to introduce numerical analysis with deep neural networks. Traditional function classes used in numerical analysis include polynomials, trigonometric polynomials, splines, finite elements, wavelets, and kernels. Deep neural networks were recently employed in numerical analysis as a class of approximation functions, demonstrating advantages over traditional function classes. These talks will cover the following topics:

  1. Deep neural network representation of a function
  2.  Optimization problems that learn a neural network
  3. Adaptive solutions of integral equations with deep neural networks
  4. Adaptive solutions of partial differential equations with deep neural networks.

Schedule

September 12nd, 9.30-11.30 Aula A
September 13nd, 9.30-11.30 Aula A

 Scritto da in 21 Agosto 2024  Senza categoria  Commenti disabilitati su Seminar: Numerical Analysis with Deep Neural Networks
Lug 222024
 

QUBO and Quantum Annealing

Prof. Michele Marchesi
University of Cagliari and NetService spa

Abstract

The seminar, lasting about an hour, presents the problems of unconstrained binary quadratic optimization, where the variables assume binary values (0/1 or -1/1), and the function to be optimized is a quadratic form with real coefficients. It will discuss various real problems that can be represented as QUBO and how to incorporate constraints using penalty coefficients. Exact classical solvers, which can only be used for small problems due to the NP-complete complexity of QUBO problems, and the main heuristic solvers: Tabu Search and Simulated Annealing, will then be presented. Finally, the Quantum Annealing approach for solving this type of problem on specialized quantum computers will be presented.

Schedule

July 25th, 9:30-10:30 (Palazzo delle Scienze, Aula B)

 Scritto da in 22 Luglio 2024  Senza categoria  Commenti disabilitati su Seminar: QUBO and Quantum Annealing
Lug 032024
 

Research in Blockchain and Quantum Technologies

Dr. Ernestas Filatovas
Vilnius University, Lithuania

Abstract

Blockchain and Quantum technologies are among the most groundbreaking advancements, attracting significant attention from industry, government, and academia. This talk highlights the research advances of our “Blockchain and Quantum Technologies Group” in both fields. In the first part of the talk, we introduce Blockchain technology, covering its main concepts such as decentralization, consensus protocols, transaction flow, etc. These key concepts later are summarized within a layered structure. We then present our recent research, including a systematic review and empirical analysis of blockchain simulators, a multi-criteria decision-making (MCDM) framework for selecting consensus protocols, a data-driven classification of consensus protocols using machine learning, and an empirical analysis of wealth decentralization in blockchain networks. This part concludes with an introduction to our novel blockchain-based platform designed to enhance research reproducibility in machine learning. The second part of the talk shifts to Quantum Computing, beginning with an overview of the current state of this technology and its potential applications across various industries. We then highlight our recent achievements, such as the development of more efficient quantum circuits for integer division and the implementation of a quantum blockchain based on hypergraphs. The talk finishes with a presentation of our ongoing research, where we propose an improved quantum annealing method to scale vehicle routing problems.

Schedule

June 20th, 10:00-12:00 (Palazzo delle Scienze, Aula Magna Matematica)

 Scritto da in 3 Luglio 2024  Senza categoria  Commenti disabilitati su Seminar: Research in Blockchain and Quantum Technologies
Lug 032024
 

From Theory to Practice: Derivative-Free Optimization, Bilevel Problems, and Real-World Applications

Prof. Remigijus Paulavičius
Vilnius University, Lithuania

Abstract

This talk explores the journey from theory to practice in derivative-free optimization (DFO), primarily focusing on the impact of DIRECT-type algorithms and their application to practical problems. While focusing on this key aspect, the presentation also delves into several other pertinent areas of mathematical optimization, such as bilevel optimization, offering a broader perspective on the field’s advancements and challenges. The presentation begins with an overview of the DIRECT algorithm’s role in DFO, highlighting its strengths and limitations. The subsequent discussion delves into advancements in DIRECT-type algorithms and their integration into software tools, facilitating practical applications. The talk presents a comparative analysis of deterministic and stochastic DFO methods through benchmarking studies, evaluating their performance and suitability for various problem domains. Then, the author showcases his contributions to the bilevel optimization field, where he worked on a general bilevel algorithm and underscored the critical need for such algorithms and software tools in practical applications, including ML. Finally, the talk showcases successful collaborations between academia and industry, highlighting the practical implementation of DFO techniques in real-world scenarios with examples from GlobeTrott Travel and Girteka Logistics, demonstrating the impact of DFO in solving complex optimization problems in business settings.

Schedule

June 19th, 10:00-12:00 (Palazzo delle Scienze, Aula Magna Matematica)

 Scritto da in 3 Luglio 2024  Senza categoria  Commenti disabilitati su Seminar: From Theory to Practice: Derivative-Free Optimization, Bilevel Problems, and Real-World Applications
Giu 172024
 

Introduction to scientific Python programming

Dr. Tamás László Storcz
University of Pécs, Hungary

Outline

  1. Scientific data and data science, working with Python ecosystem
  2. Data collection, preparation and cleaning
  3. Data visualization
  4. Feature engineering
  5. Creating and validating AI models
  6. Searching model parameters
  7. Practical data management
  8. Examples of application

Schedule

Please register to the course through this form.

June 24th, 15.00-17.00 Aula F
June 25th, 15.00-17.00 Aula F
June 26th, 15.00-17.00 Aula F
June 27th, 9.30-11.30 Aula F
June 28th, 15.00-17.00 Aula F (final test)

Exam

The final exam consists in a test, which will be taken on the last day of the course.

contatti | accessibilità Università degli Studi di Cagliari
C.F.: 80019600925 - P.I.: 00443370929
note legali | privacy