Hello there! I’m Kenneth Verstraete, a PhD researcher at the KU Leuven in Belgium.

Research

My current research topics include:

• Main projects:
  • Predicting individual treatment response using the data from randomized controlled trials: facing the fundamental problem of causal inference, I’m researching models to predict treatment effects and I’m exploring validation methodologies since the standard methods cannot be applied in this framework.
  • COPD phenotyping: using data from large COPD patient cohorts, I try to extract phenotypes based on baseline parameters, CT measures and spirometry in this heterogeneous complex disease.
• Side projects:
  • Classification in pulmonary hypertension: develop a machine-learning model that can distinguish patients with pulmonary arterial hypertension from patients with pulmonary hypertension due to left-heart disease.
  • Prediction of exacerbations in COPD based on physical activity: use the data from wearable sensors to detect patterns in physical activity that can predict acute exacerbations before their onset.

Other research interests

• Tensors: with these high-dimensional mathematical structures, interpretable patterns can be extracted without any dimensionality reduction.

Master’s theses

To obtain my Master in Engineering Science: Computer Science (2017), I performed a master’s thesis under the supervision of prof. Nick Vannieuwenhoven at the Numerical Analysis and Applied Mathematics (NUMA) research unit of KU Leuven. During this thesis, I explored tensor decompositions (the Tucker decomposition in particular) to extract features from the famous MNIST dataset to classify handwritten digits.

For the completion of my advanced Master in Artificial Intelligence (2019), I continued my work on tensor decompositions and explored the use of the canonical polyadic decomposition to extract patterns in soccer player data from the FIFA video games. I was mentored by Tom Decroos under the supervision of prof. Jesse Davis from the DTAI research group and prof. Nick Vannieuwenhoven from the NUMA research unit.