Projects
Learning and decision theory in relational dynamic domains. KU Leuven
Interacting Particle Networks: a new deep learning approach to molecular simulation of condensed phases. Ghent University
Force fields are computationally very efficient, yet coarse approximations to the potential energy surface felt by nuclei in molecules. In this project, recent breakthroughs in machine learning will be exploited to increase their reliability. The goal of this work, is to establish force fields with a novel deep learning concept, designed to “understand” many-body interactions: the Interacting Particle Network (IPN).
An exploratory study of machine learning techniques for imaging flow cytometry Ghent University
This project aims to analyse imaging flow cytometry data with advanced computational techniques. Imaging flow cytometry is a novel technology which allows making photographs of millions of individual cells in a high-throughput fashion, capturing their size and shape as well as information about protein expression. These measurements can be used to study and understand complex biological systems, such as the immune system, made up of dozens of ...
Mining of large-scale single-cell data with deep learning and parallel multi-instance algorithms Ghent University
Single-cell technologies have greatly advanced along the last few decades. The amount of singlecell
data being generated grows exponentially, in multiple areas and formats such as high content
imaging, flow or mass cytometry or RNA sequencing. However, most current machine learning
techniques cannot cope with that huge amount of information, the most commonly followed
approach being summarizing the data of all the ...
Learning Model Constraints for Structured Prediction KU Leuven
Structured output prediction based on discriminatively trained probabilistic graphical models is a powerful framework that has lead to a large improvement in predictive systems. These models, however, often require strong a priori constraints to guarantee tractable inference procedures. These constraints can limit the power of the model to provide good predictions, and can therefore be viewed as a necessary evil. This project will develop ...
Public health decision making with stochastic individual-based models: a translational framework driven by advances in health economics, model inference and reinforcement learning (ACCELERATE) Hasselt University
Machine learning for fraud analytics KU Leuven
Fraud remains a major challenge for businesses. The Association of Certified Fraud Examiners (ACFE) estimates that a typical organization loses 5% of its revenues due to fraud. Furthermore, fraudsters continuously adapt their techniques in response to fraud detection efforts, creating a need for adaptive fraud detection systems. Given the abundant availability of data, machine learning techniques seem well suited to tackle this problem. ...
Robust Directed Acyclic Graph Learning for Causal Modeling. University of Antwerp
A quantitative evaluation of deep learning techniques for unsupervised and supervised analysis of high-dimensional cytometry data Ghent University
Recent advances in cytometry allow scientists to measure many parameters at a single-cell resolution, and this for millions of cells across tens to hundreds of patients, possibly at different time points. To make sense of all this data, novel machine learning approaches for visualisation, automated population identification and subsequent differential analysis between groups of patients will be explored and compared. In this project in ...