Lectures

Lecture 1. “Multimodal Foundation Models” 1/4

Abstract TBA

Lecture 2. “Multimodal Foundation Models” 2/4

Abstract TBA

Lecture 3. “Multimodal Foundation Models” 3/4

Abstract TBA

Lecture 4. “Multimodal Foundation Models” 4/4

Abstract TBA

Gabriel Barth-Maron

Lecture 1. “Cluster-detection Methods in Network-based Data Analysis” 1/2

Cluster analysis is an important task arising in network-based data analysis. Perhaps the most natural model of a cluster in a network is given by a clique, which is a subset of pairwise-adjacent nodes. However, the clique model appears to be overly restrictive in practice, which has led to introduction of numerous models relaxing various properties of cliques, known as clique relaxations. This talk focuses on a systematic cluster analysis framework based on clique relaxation models.

Lecture 2. “Continuous Approaches to Cluster-Detection Problems in Networks ” 2/2

We discuss continuous formulations for several important cluster-detection problems in networks. More specifically, the problems of interested are formulated as quadratic, cubic, or higher-degree polynomial optimization problems subject to linear constraints. The proposed formulations are used to develop analytical bounds as well as effective algorithms for some of the problems. Moreover, a novel hierarchy of nonconvex continuous reformulations of optimization problems on networks is discussed.

Sergiy Butenko

Lecture 1. “Introduction to Foundation Models” 1/5

Abstract TBA

Lecture 2. “Large Language Models” 2/5

Abstract TBA

Lecture 3. “Efficient Language Models” 3/5

Abstract TBA

Lecture 4. “Foundation-Model Agents” 4/5

Abstract TBA

Lecture 5. “Executing Data Science Projects with Foundation Models” 5/5

Abstract TBA

Sven Giesselbach

Lecture 1. “Sequence Models 1” 1/4

Abstract TBA

Lecture 2. “Sequence Models 2” 2/4

Abstract TBA

Lecture 3. “Sequence Models 3: Liquid Neural Networks” 3/4

Abstract TBA

Lecture 4. “Sequence Models 4: Liquid Neural Networks” 4/4

Abstract TBA

Ramin Hasani

Lecture 1. “Diffusion Models” 1/3

Abstract TBA

Lecture 2. “Diffusion Models” 2/3

Abstract TBA

Lecture 3. “Diffusion Models” 3/3

Abstract TBA

Chin-Wei Huang

Lecture 1. “Large Language Model Evaluation” 1/4

Abstract TBA

Lecture 2. “Knowledge and Reasoning in Large Language Models” 2/4

Abstract TBA

Lecture 3. “Retrieval-augmented Language Models” 3/4

Abstract TBA

Lecture 4. “Tool-use in Language Models ” 4/4

Abstract TBA

Nora Kassner

Lecture 1. “Machine Learning for Modeling and Control of Complex Systems” 1/2

Abstract TBA

Lecture 2. “Machine Learning for Modeling and Control of Complex Systems” 2/2

Abstract TBA

Lecture 3. “Solving forward and inverse problems with and without neural networks”

Abstract TBA

Lecture 4. “Bayesian Uncertainty Quantification”

Abstract TBA

Petros Koumoutsakos

Lecture 1. “Introduction to Natural Language Processing (NLP) up to LLMs and associated challenges” 1/4

Abstract TBA

Lecture 2. “Identifying changes in longitudinal user generated content (I)- (recurrence and path signatures)” 2/4

Abstract TBA

Lecture 3. “Identifying changes in longitudinal user generated content (II)-(transformer based methods)” 3/4

Abstract TBA

Lecture 4. “Timeline extraction, Timeline summarisation and demo on identifying longitudinal changes” 4/4

Abstract TBA

Maria Liakata

Lecture “Twin Support Vector Machines: Advances and Applications”

The Twin Support Vector Machine (TSVM) is a powerful extension of the conventional Support
Vector Machine (SVM) algorithm, designed to address classification tasks in real world data sets.
Developed as an enhancement to traditional SVMs, TWSVM offers improved robustness and
efficiency, making it a compelling choice for various machine learning applications.

In this lecture, we delve into the theoretical foundations and practical implications of the Twin
Support Vector Machine. We begin by elucidating the fundamental concepts behind SVMs and the
motivation for the development of TWSVM. We explore the key principles underpinning TWSVM,
including the formulation of the twin optimization problem and the incorporation of twin constraints
for enhanced classification performance.

Furthermore, we delve into the algorithmic intricacies of TWSVM, elucidating its training procedure,
kernelization techniques, and model evaluation methods. We highlight how TWSVM effectively
addresses the challenges posed by high-dimensional datasets, thereby enhancing its applicability
across diverse real-world scenarios. Moreover, we investigate recent advancements and extensions
of TWSVM, particularly focusing on optimization techniques that have been developed to further
improve its performance and scalability.

Through this lecture, participants will gain a comprehensive understanding of the Twin Support
Vector Machine and its significance in modern machine learning research and applications. We aim
to equip attendees with the knowledge and insights necessary to leverage TWSVM effectively in
their data analysis endeavors, fostering innovation and advancement in the field of computational
intelligence.

References:
1. Moosaei, Hossein, Fatemeh Bazikar, Milan Hladík, and Panos M. Pardalos. "Sparse least-
squares Universum twin bounded support vector machine with adaptive Lp-norms and feature
selection." Expert Systems with Applications (2024): 123378.
https://doi.org/10.1016/j.eswa.2024.123378

2. Moosaei, Hossein, Fatemeh Bazikar, and Panos M. Pardalos. "An improved multi-task least
squares twin support vector machine." Annals of Mathematics and Artificial
Intelligence (2023): 1-21.
https://link.springer.com/article/10.1007/s10472-023-09877-8

www.ise.ufl.edu/pardalos

Panos Pardalos

Lecture “Challenges, approaches, needs and points of view for industry grade AI”

Abstract TBA

Raniero Romagnoli

Lecture “Artificial Intelligence: Where We Are, Where We Are Going?”

At present, there is a great deal of confusion as to the final objective of AI. Some see Artificial General Intelligence as the ultimate and imminent goal suggesting that it can be achieved through machine learning and its further developments.

We argue that despite the spectacular rise of AI, we still have weak AI that only provides building blocks for intelligent systems, mainly intelligent assistants that interact with users in question-answer mode.
A bold step toward human-level intelligence would be the advent of autonomous systems resulting from the marriage between AI and ICT envisaged in particular by the IoT. In this evolution, the ability to guarantee the trustworthiness of AI systems – reputed to be “black boxes” very different from traditional digital systems – will determine their degree of acceptance and integration in critical applications.

We review the current state of the art in AI and its possible evolution, including:

Avenues for the development of future intelligent systems, in particular autonomous systems as the result of the convergence between AI and ICT;
The inherent limitations of the validation of AI systems due to their lack of explainability, and the case for new theoretical foundations to extend existing rigorous validation methods;
Complementarity between human and machine intelligence, which can lead to a multitude of intelligence concepts reflecting the ability to combine data-based and symbolic knowledge to varying degrees.

In light of this analysis, we conclude with a discussion of AI-induced risks, their assessment and regulation.