Projects
Data-adaptive causal inference under weak assumptions Ghent University
Whilst data-adaptive methods (e.g. statistical learning) have been successful in prediction problems, statisticians have been cautious in adopting them for inferring causal effects. This is because these techniques are prone to ‘plug-in’ bias and drastically complicate inference. However, recent work has shown how data-adaptive methods can be used in a first step for evaluating the effect of an intervention, whilst still yielding valid ...
ANUBIS: Aligned oNline and multilevel User and entity Behavior KU Leuven
Fraud is a fierce threat to digital business. A typical organization is estimated to lose 5% of its revenues due to fraud, which is hard to eradicate since dynamic, system-dependent and organizationspecific. Powerful and intelligent fraud detection systems are therefore of crucial importance, to timely block, prevent and contain fraud and to mitigate losses. User and entity behavior analytics essentially profile the activity of users, peer ...
Improving the prediction of language recovery in stroke patients by including risk and protective neurocognitive factors. KU Leuven
Aphasia is a language disorder most commonly caused by a cerebrovascular accident, or stroke, in the left hemisphere. Approximately 15-45% of patients in acute stroke settings have aphasia, which impairs their communication and dramatically affects their quality of life. In the first months after stroke, there is often spontaneous and intervention-induced recovery, but for 26-43% a chronic language deficit remains. To date, individual ...
The role of auditory-motor speech processing and higher cognitive abilities in language acquisition Ghent University
Children appear to be better language learners than adults. The precise reason why remains unclear. Language learning is complex and involves unconsciously picking up statistical regularities from continuous streams of speech sounds, such as word boundaries or rules on word order. Our previous work shows that higher cognitive abilities interfere with statistical learning processes in adults. This provides a plausible first explanation for why ...
Machine-learning-enhanced identification of electron transporting biomolecules in electroactive microorganisms using combined Timeof- Flight Secondary Ion Mass Spectrometry (ToF-SIMS) and Scanning Probe Microscopy (SPM) data. Hasselt University
Machine-learning-enhanced identification of electron transporting biomolecules in electroactive microorganisms using combined Timeof- Flight Secondary Ion Mass Spectrometry (ToF-SIMS) and Scanning Probe Microscopy (SPM) data. Hasselt University
Behavioural and neural processes underlying the formation of phoneme representations in children with dyslexia. KU Leuven
Learning to read consists of coupling letters to speech sounds (i.e. phonemes). In this perspective, it is important to built up, already before reading, well specified representations of phonemes. The development of good phoneme representations is based on good auditory skills, since it requires to hear subtle differences between phonemes. Above, sensitivity to the frequency distribution of speech sounds is needed to learn to categorize ...
Provable Safe Learning-based Control of Uncertain Systems KU Leuven
When it is designed well, a controller is invisible and requires no human involvement after deployment. This hinges on two crucial qualities: certifiable safety and adaptiveness. Control theory addresses safety by ensuring stability, constraint satisfaction, etc. by imposing assumptions on the dynamical system. However, this often restricts adaptability -- a core trait of Machine Learning (ML) and Reinforcement Learning (RL), which excel in ...