Title Promoter Affiliations Abstract "Identification of adaptive mechanisms leading to reduced antibiotic susceptibility in bacterial biofilms using experimental evolution and machine learning approaches" "Serge Van Calenbergh, Willem Waegeman, Filip Van Nieuwerburgh, Tom Coenye" "Department of Data analysis and mathematical modelling, Department of Pharmaceutics, Department of Pharmaceutical analysis" "Because many mechanisms of reduced sensitivity in bacterial biofilms are still unknown, it is impossible to predict resistance. In this project we will allow bacteria to evolve in vitro in the presence of antibiotics, in order to map all mutations, differences in gene expression and relevant phenotypic characteristics. This will allow to develop a prediction algorithm using machine learning." "Retrokit - Optimising Machine Retrofitting with Retrokit: A Modular Machine Learning Approach for Edge-Based Condition Monitoring" "Jonas Lannoo" "Smart Technologies, KU Leuven campus Brugge (KULAB), Iot & data driven solutions" "An increasing number of companies are transitioning to Industry 4.0, yet facing an aging fleet of machines. Retrofitting is a cost-effective and non-invasive strategy to make legacy industrial machinery compliant with Industry 4.0 standards by incorporating additional technological features. The M-Group at KU Leuven Bruges Campus has created a condition-monitoring methodology using unsupervised anomaly detection with machine learning for retrofitting applications. This resulting model is about 50% smaller than the state of the art, yet as performant in terms of fault detection accuracy. This reduced size is an essential advantage when compared to competing approaches, as it enables to perform the actual fault detection on Micro-Controller Units (MCU) at the edge. This project aims to develop the Retrokit, a flexible and modular condition monitoring system that encapsulates the resulting methodology in an embedded device. As such, it will include generic components for data collection, processing, and classification, focusing on a minimally invasive implementation through the analysis of electrical signal data. The objective is to develop a modular design that can be connected to fit the specific application. The M-Group will engage in a close partnership with the IoT Lab at VIVES University of Applied Sciences to leverage their hardware expertise in the field of embedded devices." "Retrokit - Optimising Machine Retrofitting with Retrokit: A Modular Machine Learning Approach for Edge-Based Condition Monitoring" "Mathias Verbeke" "Declarative Languages and Artificial Intelligence (DTAI), Distributed and Secure Software (DistriNet), Waves: Core Research and Engineering (WaveCore)" "An increasing number of companies are transitioning to Industry 4.0, yet facing an aging fleet of machines. Retrofitting is a cost-effective and non-invasive strategy to make legacy industrial machinery compliant with Industry 4.0 standards by incorporating additional technological features. The M-Group at KU Leuven Bruges Campus has created a condition-monitoring methodology using unsupervised anomaly detection with machine learning for retrofitting applications. This resulting model is about 50% smaller than the state of the art, yet as performant in terms of fault detection accuracy. This reduced size is an essential advantage when compared to competing approaches, as it enables to perform the actual fault detection on Micro-Controller Units (MCU) at the edge. This project aims to develop the Retrokit, a flexible and modular condition monitoring system that encapsulates the resulting methodology in an embedded device. As such, it will include generic components for data collection, processing, and classification, focusing on a minimally invasive implementation through the analysis of electrical signal data. The objective is to develop a modular design that can be connected to fit the specific application. The M-Group will engage in a close partnership with the IoT Lab at VIVES University of Applied Sciences to leverage their hardware expertise in the field of embedded devices." "Learning Invariant Models in a Causal Machine Learning Framework." "Internet Data Lab (IDLab)" "Traditional machine learning techniques focus on developing predictive models that have the sole purpose of obtaining a high degree of accuracy on a given data set. These types of models exploit any type of association between the input and target variables that may increase the performance. However, in practice, the training and test distribution often differ significantly, resulting in unreliable and failing models. The key to learning generalizable models that work in a broad range of environments (and that are not affected by small changes in the test distribution) lies in learning causal predictive features. However, learning causal models under changing environments and in systems with hidden confounders is an unsolved problem and is directly connected to the generalisation gap. In this project, we aim to use the novel framework of causal machine learning to develop algorithms that can handle changing environments. More specifically, this project focuses on learning invariant and causal representations from data using causal machine learning. The results are models that are proven to be more generalizable, can cope with interventions, and are able to extract interpretable causal relations directly from data." "Declarative modeling for machine learning and data mining." "Luc De Raedt" "Declarative Languages and Artificial Intelligence (DTAI)" "Declarative models specify what conditions need to satisfied in order to obtain a solution to a specific problem. Declarative models contrast with the traditional procedural approach which specify how such solutions must be computed. A declarative modeling paradigm will be developed and applied to the areas of machine learning, data mining and experimentation. The declarative modeling paradigm that we will pursue consists of three key components: A modeling language (M language) which is a high level declarative language for specifying the relevant domain knowledge, independent of a particular task (M component). A solver which accepts input in a more low-level task-oriented language (S language) and performs a particular computational task (S component). A programming platform in which a user employs a general purpose language to solve a specific computational task. Today, there exist no general declarative approaches to machine learning, data mining and experimentation. Therefore, contrasting these domains with contemporary approaches to declarative modeling (pursued in knowledge representation and constraint programming) forms an ideal setup for realizing breakthroughs in declarative modeling as well as in machine learning, data mining and experimentation." "Principled Machine Learning for longitudinal biomedical data with repeated measures" "Celine Vens" "Public Health and Primary Care, Kulak Kortrijk Campus" "Advanced computational methods using machine learning could become an indispensable tool for medical practitioners in a multitude of clinical settings like prognostic modeling and patient risk stratification. However, most Machine Learning techniques were not designed to account for the complexities stemming from the longitudinal and otherwise hierarchical nature of most health datasets. Not only do correlations between repeated measurements within patients violate the fundamental assumption of most algorithms that all observations are independent and identically distributed, but most methods are also incapable of handling missing values which arise in most longitudinal datasets due to irregular follow-up times or patient dropout. Imputing these values using Machine Learning without accounting for the mechanism by which those missing values came to be could lead to biased predictions or inferences. The black-box nature of these methods also makes it challenging to not only explain certain predictions about individual patients, but also draw rigorous inferences on the impact of specific variables and disentangle whether observed differences are statistically significant or due to chance.The main objectives of this project are developing Machine Learning methods for working with longitudinal datasets which are grounded in sound statistical theory and leveraging their ability to model complex non-linear relationships in a principled manner. Special attention will be given to clinical applications in Ophthalmology." "Machine Learning @ the Extreme Edge" "Joeri Verbiest" "Karel De Grote Hogeschool, Research Centre Sustainable Industries" "In Europe, embedded intelligence is one of the critical technologies and Artificial Intelligence (AI) is a strategic technology. For Flanders, one of the challenges is real-time and energy-efficient information extraction and processing at the edge, ""the edge,"" using AI. Currently, this is mainly done with ""intelligent"" edge systems. Very recently, there is a trend to perform machine learning (ML) on end-point devices. These devices are located at the extreme edge, at the boundary between the analog, physical world, and the digital world. They consist of one or more sensors and an embedded resource constrained device, a device with a limited amount of memory, computing power, and energy consumption. The challenge is to develop accurate, energy-efficient ML models. This can be accomplished through tiny machine learning (tinyML), a sub-field in machine learning. AI has applications in many sectors, such as healthcare. In the future, if we want to move to low-cost and convenient patient monitoring, we need wireless and battery-powered systems, working in a portable wireless network, that collectively make predictions via ML. Consequently, healthcare is one of the domains in which tinyML can play an important role. The research project looks at how end-point devices and tinyML technology can be used in the development of an intelligent portable monitoring system. A system that can be deployed within healthcare. Work is being done through a use case, remote monitoring of a rehabilitation process, real-time monitoring of movement (step pattern, movement, activity, ...) in patients in a home environment. Such systems, among other things, allow rehabilitation to be better tailored to the individual needs of the patient. The possibilities of tinyML technology are demonstrated. A technology that allows developers to develop intelligent devices. In addition, the requirements of a future patient monitoring system are examined from the user's point of view. With this knowledge, developers can better adapt the portable intelligent monitoring systems to the requirements of the end users." "Machine Learning Acceleration on Heterogeneous Platforms" "Josep Balasch Masoliver" "Dynamical Systems, Signal Processing and Data Analytics (STADIUS)" "Advances in the fields of biomedical sensors, wearables and medical implants, in combination with state-of-the-art algorithms from signal processing, machine learning and artificial intelligence, are transforming the healthcare landscape. Systems built around these technologies enable remote health monitoring, improve patient care, detect life-threatening conditions or even predict health events. Yet the full integration of such technologies into embedded/edge platforms poses several challenges: edge platforms are by nature resource-constrained and power/energy-limited, while data processing algorithms are computational and memory demanding. Managing this trade-off demands research on efficient implementation methods tailored to the system requirements. In this context, the proposed PhD vacancy aims to investigate and develop digital architectures that support local data processing at edge nodes. The idea is to leverage on modern heterogeneous platforms which combine embedded processor cores (software) with reconfigurable FPGA logic (hardware). The flexibility of such platforms offers better support for evolving algorithms, reduced numerical precision or custom scheduling techniques, necessary to tailor the implementation to the system requirements. As part of the work, the candidate will explore modern trends in FPGA technologies which facilitate the flow for implementing complex machine learning frameworks by working at higher abstraction levels." "Structured Machine Learning for Mapping Natural Language to Spatial Ontologies" "Marie-Francine Moens" "Informatics Section" "Natural language understanding is one of the fundamental goals of artificial intelligence. An essential function of natural language is to talkabout the location, and translocation of objects in space. Understanding spatial language is important in many applications such as geographical information systems, human computer interaction, the provision of navigational instructions to robots, visualization or text-to-scene conversion, etc.Due to the complexity of spatial primitives and notions, andthe challenges of designing ontologies for formal spatial representation, the extraction of the spatial information from natural language stillhas to be placed in a well-defined framework. Machine learning has not systematically been applied to the task, and no established corpora are available. In this thesis I study the problem from cognitive, linguistics and computational points of view, with a primary focus on establishinga supervised machine learning framework.This thesis makes five mainresearch contributions. The first is the design of a spatial annotationscheme  to bridge between natural language and formal spatial representations. In this scheme the universal and commonly accepted cognitive spatial notions and multiple well-known qualitative spatial reasoning models are applied.The second is the definition of a novel computational linguistic task that utilizes the annotation scheme to map natural language to spatial ontologies. For this task I have built  rich annotated corpora and an evaluation scheme.The third is a detailed investigation of the linguistic features and structural characteristics of spatial language that aid the use of machine learning in extracting spatial roles and relations from annotated data. The learning methods used are discriminative graphical models and statistical relational learning.The fourth is the proposal of a unified structured output learning model for ontologies. The ontology components are learnt while taking intoaccount the ontological constraints and linguistic dependencies among the components. The ontology includes roles and relations, and multiple formal semantic types. The fifth is the proposal of an efficientinference approach based upon constraint optimization. It can deal witha large number of variables and constraints, and makes building a global structured learning model for ontology population, feasible. To test the approach I have performed an empirical investigation using my spatialontology.The application of my proposed unified learning model for ontology population is not limited to the extraction of spatial semantics, it could be used to populate any ontology. I argue therefore that this work is an important step towards automatically describing text with semantic labels that form a structured ontological representation of the content.    " "AI Musicking: Innovative Approaches to Musical Co-creation through Machine Learning" "Umut Eldem" Creatie "Over the last few years, the concept of AI & ML has established itself not only as an experimental field of research, but also as a toolset applicable to creative use. While there have been breakthroughs and public adoption in the fields of text-based generation (ChatGPT1) and image-based generation (DALL-E2), the existing tools for musical generation have yet to see widespread adoption in terms of public use and artistic applications. Using data in form of audio recordings and musical mappings, AI algorithms are able to learn, process, and reproduce patterns and stylistic properties of complex musical systems without the need to explicitly formulate methods for generating them.3 As the technological aspects of such AI generation is quickly evolving and is under close scrutiny, an artistic methodology and framework relating to the use of AI & ML in terms of contemporary musical human-machine co-creative practices have yet to be thoroughly explored.4 This is a field that needs to be investigated on the basis of the current and potential future impact of such tools in the aforementioned practices. The proposed project aims to investigate and develop specific artistic methodologies regarding using musical AI tools in the human-machine co-creative process through studying and applying the data generated by the artist-researchers of the CREATIE research group. Several artist-researchers of CREATIE are now involved in the process of applying machine learning in their creations with the need to improve and develop new approaches, methodologies and skills. onderzoeksvragen: 1. Which are the skills artist-researchers in music should develop to co-create with AI? 2. In which ways can data be generated and processed by the existing Machine Learning tools with the aim of training a creative tool? 3. Which current and future frameworks and methodologies are suitable for the musical co-creative process between human and AI? 4. What are the limits and potentials of using AI to generate music? 5. Which are authorship’s issues and sense of agency in a machine learning co-creation? 6. What is the ethics of the data generated by the existing activities of researchers?"