Titel Deelnemers "Korte inhoud" "Deep kernel principal component analysis for multi-level feature learning" "Panos Patrinos, Johan Suykens" "Principal Component Analysis (PCA) and its nonlinear extension Kernel PCA (KPCA) are widely used across science and industry for data analysis and dimensionality reduction. Modern deep learning tools have achieved great empirical success, but a framework for deep principal component analysis is still lacking. Here we develop a deep kernel PCA methodology (DKPCA) to extract multiple levels of the most informative components of the data. Our scheme can effectively identify new hierarchical variables, called deep principal components, capturing the main characteristics of high-dimensional data through a simple and interpretable numerical optimization. We couple the principal components of multiple KPCA levels, theoretically showing that DKPCA creates both forward and backward dependency across levels, which has not been explored in kernel methods and yet is crucial to extract more informative features. Various experimental evaluations on multiple data types show that DKPCA finds more efficient and disentangled representations with higher explained variance in fewer principal components, compared to the shallow KPCA. We demonstrate that our method allows for effective hierarchical data exploration, with the ability to separate the key generative factors of the input data both for large datasets and when few training samples are available. Overall, DKPCA can facilitate the extraction of useful patterns from high-dimensional data by learning more informative features organized in different levels, giving diversified aspects to explore the variation factors in the data, while maintaining a simple mathematical formulation." "Federated learning under a digital communication model" "Panos Patrinos, Marc Moonen" "Distributionally Robust Model Predictive Control for Safety-Critical Systems: with Applications in Autonomous Driving" "Mathijs Schuurmans" "The availability of reliable methods for data-driven, optimal decision making under uncertainty is of central importance in several fields of study, including operations research, control, machine learning, and many more specialized application domains. Current (black-box) machine learning techniques, despite their many successes, still provide very few guarantees regarding correctness and safety, impeding their adoption for many high-stakes and safety-critical applications. This drawback can be partially mitigated by further increases in computational power and dataset sizes, but nevertheless, such a remedy is insufficient for applications akin to robotics, where computational hardware or availability of informative data may be limited. This thesis aims to address this shortcoming by developing novel methodologies for data-driven decision-making and control, which are applicable in online settings, and which provide guarantees regarding performance and constraint satisfaction for finite sample sizes. To this end, distributionally robust optimization (DRO) serves as our main tool, as it provides many of these desirable guarantees by design. Furthermore, its conceptual simplicity makes it a versatile approach for many specific problems. In particular, we highlight the following contributions: (i) We present distributionally robust techniques for synthesis of stabilizing controllers for Markov jump linear systems, where the distribution of the switching process is unknown; (ii) We develop a theoretical framework for distributionally robust model predictive control of general (nonlinear) Markov jump systems (also referred to as Markov switching systems). This gives rise to multi-stage risk-averse optimization problems, including nested risk measures in both the costs and the constraints; (iii) In order to efficiently solve these problems, we propose tractable reformulations and a tailored, massively parallellisable solver; (iv) We propose a novel DRO method, called Cost-Aware Distributionally Robust Optimization (Cadro), which, by exploiting the structure of the cost function in the design of its ambiguity set, results in less conservative solutions to data-driven DRO problems, while retaining the same guarantees; and (v) We apply our developments to case studies within automated driving, and demonstrate empirically the effects of relevant design trade-offs in these examples." "Dualities for Non-Euclidean Smoothness and Strong Convexity under the Light of Generalized Conjugacy" "Emanuel Laude, Panos Patrinos" "The “Eagle” Approach To Train Electrical Engineers With Collaborative Problem-Solving Skills" "Merijn Van Deyck, Martijn Deckers, Abdul Saboor, Pouya Mehrjouseresht, Zhenda Zhang, Arne Symons, Alexander Bodard, Marian Verhelst, Alexander Bertrand, Ruth Vazquez Sabariego, Panos Patrinos" "Engineering education plays a critical role in addressing the ever-increasing environmental and societal challenges, and collaborative problem solving (CPS) is a vital skill for engineers to tackle such complex multidisciplinary challenges and develop high-quality solutions. The EAGLE project at KU Leuven exemplifies CPS implementation in electrical engineering education, providing students with real-world connections and deep learning opportunities to develop teamwork, problem-solving, and negotiation skills. This paper presents the development and implementation of EAGLE, a year-long hands-on, multidisciplinary challenge in which teams of 10-12 students design and develop an autonomous drone capable of flying to a remote landing station. It focuses on the project organization, innovative coach-based teaching and grading system, and the multi-dimensional evaluation and grading processes employed. The insights gained from the EAGLE project can offer valuable lessons for future project-based learning initiatives and encourage the adoption of innovative teaching and learning approaches in engineering education. By sharing our experiences, we aim to inspire other educators to integrate real-world projects into their curricula, emphasizing the significance of hands-on learning, teamwork, and CPS in engineering education." "Model-free decentralized training for deep learning based resource allocation in communication networks" "Panos Patrinos, Marc Moonen" "Safe, Learning-Based MPC for Highway Driving under Lane-Change Uncertainty: A Distributionally Robust Approach" "Mathijs Schuurmans, Panos Patrinos" "Anderson Accelerated Feasible Sequential Linear Programming" "David Kiessling, Alejandro Astudillo Vigoya, Pieter Pas, Panos Patrinos, Jan Swevers" "Gauss-Newton meets PANOC: A fast and globally convergent algorithm for nonlinear optimal control" "Pieter Pas, Andreas Themelis, Panos Patrinos" "SPOCK : A proximal method for multistage risk-averse optimal control problems" "Alexander Bodard, Mathijs Schuurmans, Panos Patrinos"