Projects
Explainable, human-in-the-loop hybrid AI, resolving conflicting user feedback Ghent University
One of the biggest struggles hindering adoption of AI in industry (e.g. production processes) is the fact that they see AI as a fully data driven approach, requiring a lot of data, which is often not available. Hybrid AI softens the need for big data by including domain knowledge into the AI. Whereas today’s hybrid AI solutions embed objective knowledge into the AI, here, an approach is researched that captures the tacit knowledge of ...
Explainable, human-in-the-loop hybrid AI, resolving conflicting user feedback and in a federated learning setting Ghent University
Hybrid AI softens the need for big data by including domain knowledge into the AI. Current solutions
embed objective knowledge. This research extends this by combining this with the tacit knowledge
of experienced engineers, while keeping the discrepancies between them. Uncaptured knowledge is
considered in the form of user feedback on the model outputs, including conflicting feedback between different users.
Project RESONANCE*: The development of "fair governance", a new interdisciplinary psycho-legal research approach to assess public algorithm-assisted decision making *faiRnEss Self-determination gOverNANCE. University of Antwerp
Time series analysis with the human in the loop KU Leuven
We will develop new methods for the analysis of time series that allows human to steer the analysis and interpret results (i.e., a “human-in-the-loop” approach to time series analysis). To this aim, we will investigate methods that allow for flexible supervision in time series analysis, and methods that increase the explanatory power of the analysis. Concrete topics that we will focus on include: - semi-supervised clustering of time series - ...
Interactive embedding of high-dimensional data with shape templates Ghent University
Dimensionality reduction (DR) is widely used to condense data for subsequent application of machine learning algorithms and to learn about the high-level structure of the data by providing low-dimensional embeddings used for visualization. Existing DR techniques such as t-SNE or UMAP are not guaranteed to represent the high-dimensional topological structure of the data faithfully. While topological embedding methods aim at modeling the ...
Mixed-initiative explanation methods: towards the next generation of interactive machine learning steered with rich feedback of non-expert users KU Leuven
AI for sales prediction in an environment with limited historical data VIVES
Unveiling the evolution of the physiological network during sleep KU Leuven
Balance supporting and energy-efficient control of wearable robotic devices through neuromechanical modelling KU Leuven
The control of robotic assistive devices is mainly optimized using experimental design approaches. Despite the success of this approach in reducing the energy cost of walking in healthy subjects, such benefits have not yet been shown in impaired subjects. Furthermore, the robotic device further impairs their ability to control balance, therefore restricting pathological subjects to the use of crutches. Simulation-based methods have the ...