Title Promoter Affiliations Abstract "A computational model of the usage-based acquisition of abstract constructions and grammatical categories." "Piet Desmet" "Faculty of Arts, Kulak Kortrijk Campus, Comparative, Historical and Applied Linguistics, Kulak Kortrijk Campus" "This project starts from an important insight gained from usage-based theories of language acquisition, namely that the ability of children to learn language is based on 2 general cognitive capacities: intention reading and pattern finding. Intention reading refers to the capacity of children to share attention, recognise gestures and understand the communicative intentions of their interlocutors. Pattern finding refers to the ability of children to recognise similarities and differences in their sensory-motor experiences and to use this ability for schema formation. Together, intention reading and pattern finding provide thus the necessary mechanisms for generalising across different communicative interactions, thereby constructing abstract schemata that represent the linguistic knowledge of a language user. While there is ample empirical evidence to support this hypothesis, there exist to date no faithful mechanistic models of the processes involved in acquiring such abstract schemata. The objective of this research project aims to fill this gap by providing a fully operational mechanistic model of the usage-based acquisition of abstract schemata (constructions), along with a system of grammatical categories that captures how these schemata interact with each other." "Seeing it in color: A usage-based analysis of color names in advertising." "Dirk Geeraerts" "Quantitative Lexicology and Variational Linguistics (QLVL), Leuven" "This project ties in with a specific strand of research developed withthe research group Quantitative Lexicology and Variational Lingusitics,i.e. a multivariate usage-based grammar approach. This approach triesto provide a corpus-based,statisti­cally well-tested description ofmultifactorial linguistic phenomena, i.e. phenomena that aresimultaneously determined by semantic, discursive, and variationalfactors. Specifically, in line with the lexicological specialization ofthe research team's previous studies, the present PhD project willfocus on the onomasiologicaldifferentiation between near-synonyms.In a more general perspective, this type of study may be extendedtowards near-synonymous constructions rather than lexical items.Thespecificinnovation the research group's approach brings to this field ofenquiry resides in two things: the use of advanced quantitativetechniques of analysis, and the incorporation of lectal variation as anintegrating part of the multifactorial nature of language." "It's all frequency? - testing usage-based theories of language change using agent-based models" "Dirk Speelman, Freek Van de Velde" "Quantitative Lexicology and Variational Linguistics (QLVL), Leuven" "This project proposes to use regularization methods from machine learning, more specifically Elastic Net regression (and its siblings Ridge and Lasso), to look into lexical semantic effects in morphosyntactic alternances. These regularization techniques apply shrinkage to the coefficients and can thus be used for variable selection, especially when the number of predictors is very large. In variationist studies, this is often the case if one wishes to enter lexemes associated with a construction into a regression model to predict constructional variants. We combine the Elastic Net regulator with k-fold cross-validation - a standard procedure - to avoid overfitting. Our approach mitigates the various drawbacks present in alternative approaches that are currently used in variationist linguistics, like random factors in mixed models and collostructional analysis. We look at ten multifactorially driven alternances from Dutch. The project offers a transparent pipeline that can easily be extrapolated to other case studies, and to other languages." "Form and function in usage-based construction grammar: A semantic/pragmatic analysis of clause-internal irrelevance marking in German." "Torsten Leuschner" "Department of Linguistics" "German W-immer/auch-connectives occur in universal concessive conditionals, nonspecific free relatives, and elliptical constructions. By analyzing both formal and semantic/pragmatic differences between these constructions, I will demonstrate that a more precise characterization of a construction’s function is required in order to distinguish formally similar constructions in the constructional network and explain subtle differences between them." "Image-based profiling of yeast cells to identify molecular mechanisms of mutational robustness" "Sander Govers" "Molecular Biotechnology of Plants and Micro-organisms" "Organismen hebben mutaties nodig om te evolueren, maar moeten dit balanceren met het compenseren van mogelijk negatieve effecten van deze mutaties. Bepaalde cellulaire mechanismes zouden de fenotypische gevolgen van mutaties kunnen helpen beperken en zo mutationele robuustheid verlenen. Ondanks het belang van zulke mutationele robuustheid voor het begrijpen van evolutie en genetica, blijft onze kennis hierover verrassend beperkt. Dit komt omdat het bestuderen van robuustheid enorm uitdagend is: het vereist het onderzoeken van de rol van een groot aantal genen op het effect van een groot aantal (nieuwe of bestaande) mutaties op verschillende eigenschappen en in verschillende omgevingen. Dit project combineert kwantitatieve beeldgebaseerde profilering van grote S. cerevisiae mutantenbanken met verdere functionele karakterisatie-experimenten om genen en mechanismen te identificeren die geassocieerd zijn met het verminderen van mutatie-effecten." "RePhactor - Phage-based biosynthetic gene cluster refactoring for the discovery and yield improvement of novel bioactive natural products" "Joleen Masschelein" "Molecular Biotechnology of Plants and Micro-organisms, Animal and Human Health Engineering (A2H)" "Microorganisms produce a wealth of structurally diverse bioactive natural products with important applications in medicine and agriculture. Unfortunately, the majority of natural product biosynthetic gene clusters (BGCs) are not, or only minimally, expressed under laboratory conditions. This poses a major challenge to the discovery and functional characterization of novel natural products, especially in non-model bacteria. Current methods for activating and enhancing BGC expression, such as heterologous expression in model organisms, are costly, inefficient and require extensive screening capacity. To overcome these limitations, this project aims to draw inspiration from Nature and exploit the ability of phages to hijack and control bacterial metabolic pathways. Specifically, it aims to activate and enhance the expression of natural product BGCs in non-model bacteria by integrating orthogonal, phage-derived genetic elements while taking into account the underlying regulatory mechanisms that modulate these pathways. As a proof concept, this innovative phage-based refactoring pipeline will be applied to the identification of cryptic metabolites in five diverse Pseudomonas species and to the yield enhancement of a natural product anticancer agent with therapeutic potential. Together, these innovations will have broad translational applications and will provide a much more efficient, sustainable and cost-effective means of generating industrially valuable natural products." "PhD position – Efficient Deep Learning Strategies for Image-Based Insect Recognition" "Wouter Saeys" "Mechatronics, Biostatistics and Sensors (MeBioS)" "According to FAO estimates, pests cause up to 40% of the world's crop yield to be lost each year. Over $70 billion is lost annually to invasive insects in the global economy. On a global scale, invasive species have a significant role in ecosystem services deterioration, biodiversity loss, and ecological degradation. A sensor system that allows for the rapid identification of flying insects is currently lacking, and identification of insects is mainly based on counting insects on specialized traps. This procedure is subjective and time consuming. As a result, it is only performed on a weekly basis, and at a limited number of locations. Therefore, a rapid system for insect detection would be highly valuable. In the MeBioS division, basic research has been conducted to develop insect recognition sensor devices. We specialize primarily on image and wingbeat signature analysis, with an emphasis on insect pests in the fruit and vegetable producing industry. To this end, we collaborate with several partners in Flanders and abroad to bring together all required expertise. We have already demonstrated that similar species may be identified under laboratory conditions. Now, we want to conduct research on (1) more advanced setups that could potentially improve classification results; (2) robust systems, both from a hardware and software perspective, that enable accurate measurements in the field; (3) data analytics that combine information from multiple sources; and (4) the foundation for a remotely accessible sensor network." "Deep Learning for Advanced Image-Based Semiconductor Metrology" "Jesse Davis, Stefan De Gendt" "Declarative Languages and Artificial Intelligence (DTAI), Sustainable Chemistry for Metals and Molecules" "As EUV based lithography gets adopted to keep scaling semiconductor devices in our chips, new metrology and inspection challenges arise. We need to measure these small dimensions fast but without losing accuracy and repeatability. Metrology and inspection are at the heart of process control. Without adequate metrology and inspection capability yields suffer. Conventional process and metrology tools, used in the industry, while generating a lot of data, do not always use them in feed-forward and feed-back cycles. However, there is a continuous need for better insight into process control by using this massive data. The sources of these data may range from tool process logs to lab metrology to computational to FAB metrology inspection. Manual supervision, analysis and finding any relevant inter and /or intra-correlation between these monstrous data sources is nearly impossible and therefore requires better data analysis methods and advanced machine learning techniques. The goal of this project is to use data from the manufacturing tools and use them for building models for better process control and correlate with the electrical performance of devices. The PhD candidate will learn conventional process flow and will be responsible to work collaboratively toward developing and applying “machine learning' based optimization algorithms with a goal to tackle the aforementioned challenges in terms of 1) Reducing computational cost, 2) reduce tool cycle time, 3) predictive process control approach in enabling advanced node semiconductor manufacturing, and 4) improving metrology data. Machine learning applicability includes: 1. Brainstorm “Technical diligence” of the project: to meet desired performance and engineering timeline. 2. Tool data analysis: Collect data, analyse data, and suggest hypothesis with expertise feedback loop. 3. Image processing applicability: Collect Image data (SEM/TEM/EDR/..), suggest ML based hypothesis to extract improved SEM based measurements. 4. Machine learning modelling – build from scratch or improving an existing algorithm for a given application. 5. Collaboration on patent/publications and presentations at international conferences/journals. 6. Supervision of master theses related to the subject of this PhD." "Image-based appearance metrology" "Frédéric Leloup" "Informatics Section" "The appearance impression of a product comes from its various attributes, such as the colour and gloss. For quality control purposes a comprehensive evaluation of these attributes is mandatory in several application fields and industries. To this end, standardized methods and metrics have been introduced. However, with the technical innovations e.g. made in the printing and paint industry, today new effect pigments and gonio-apparent coatings have emerged for which the conventional optical characterization methods are inadequate. In this research project, innovative measurement methods by aid of hyperspectral imaging devices will be investigated and procedures developed to determine the above-mentioned appearance characteristics. In the future, these methods should allow materials to be more accurately designed and controlled." "Development of an image-based multiparametric drug response signature to predict clinical therapy response in cancer patients from ex vivo tumoroid screenings." "Hans Prenen" "Antwerp Surgical Training, Anatomy and Research Centre (ASTARC), Center for Oncological Research (CORE)" "Precision oncology has been shown to greatly improve outcomes of cancer patients, with tailored treatment approaches that consist of patient-directed therapies on the molecular characteristics of a patient. Despite this, chemo- and radiotherapy are still the basis of most standard treatment regimens, especially for gastrointestinal (GI) cancer patients. Importantly, there are significant differences in how GI cancer patients respond to standard-of-care (SOC) chemotherapy (CT) and chemoradiation (CRT), resulting in a majority of patients experiencing either over- or undertreatment and a delay in starting the optimal treatment. Tailored treatment approaches for SOC CT/CRT to enable precision oncology for these standard therapies is of high interest in order to improve quality-of-life and survival of GI cancer patients. With no existing predictive biomarkers for CT/CRT, and genomic profiling falling short on this front, there is therefore a clear unmet medical need for a novel model that can distinguish CT/CRT responders and non-responders in GI cancer patients. Patient-derived tumor organoids (PDOs), a functional precision oncology strategy, are 3D vivo models generated from individual patient tumor tissue and have recently emerged as a promising tool for predicting CT/CRT responses in cancer patients. PDO-guided treatment has not yet been implemented in the clinic, because some limitations need to be overcome first. With this study, we aim to overcome the most important limitations by developing a multiparametric, live-cell imaging-based drug response signature for ex vivo PDO screenings that enables monitoring of the true PDO drug response. We hypothesize that this will drastically improve the predictive value of PDOs and feasibility of using PDO drug screenings in routine clinical practice. To test this and as proof-of-concept we will also perform a multicentric prospective observational cohort study with our novel PDO screening platform for prediction of neoadjuvant CT/CRT response in rectal and esophageal cancer patients in regional hospitals. If successful, we aim to set up a prospective clinical phase-1 trial in the future, and on the long term implement our PDO drug response signature as a tool to help guide clinical decision-making of CT/CRT treatment choices for GI cancer patients."