Upcoming Projects
Learning from Sparse Examples with graph learning approaches
In this project, we aim to develop advanced machine-learning methods that can deal with sparse and possibly multimodal data. The sparsity may be in the limited number of instances available or in the limited number of labels available. This is in the context of the HAICu project. Within HAICu, Artificial Intelligence researchers and Digital Humanities researchers will co-develop a scientific solution to unfold the true potential of the current heterogeneous digital heritage collections and provide easier and reliable data access to citizens, journalists, civic organizations, and other societal users.
Ongoing projects
Virtual child growth monitor
Tilburg University’s Zero Hunger Lab was tasked with developing an AI algorithm capable of measuring body parameters—such as height, circumference, and weight—from images. This algorithm can be integrated into the app to automatically detect malnutrition in children globally. (PhD candidate: Hezha Mohammedkhan)
Exploring the link between poverty, nutrition, cognitive functioning and underlying brain networks with Graph Learning
The ICON project From impoverished to enriched brains: a lifespan perspective on the link between poverty, nutrition, cognitive functioning and underlying brain networks is a condortium project bringing expertise from various disciplines PhD project aims to develop the necessary technical tools to answer the consortium’s research questions. To address the challenges of analyzing socioeconomic and brain data together, the project will create a theoretical framework using Graph Theory, Machine Learning, and Topological Data Analysis (TDA). The work involves three key components: first, developing graph data augmentation techniques to address the lack of large datasets that include both brain and socioeconomic data; second, creating multimodal data integration techniques to interpret how graph data relates to socioeconomic factors; and third, developing Graph Neural Network (GNN) models and Explainable AI (XAI) methods to utilize these approaches. (PhD candidate: Valentina Sanchez Melchor)
Explainable AI for learning methods on graph-structured data
This PhD project focuses on developing a model-agnostic explainability method for machine learning systems that process graph-structured data. To facilitate this, a symbolic representation of graphs will be designed. Different architectures of graph neural networks will be employed as the models to address the graph-learning challenge. The results will be supported by model-specific analyses to verify the robustness of the model-agnostic approach and improve the interpretability of the machine learning system. (PhD candidate: Bettina Soo ́s)
Artificial Intelligence for Power Load and Renewable Energy Forecasting in Electricity Grids
The scientific challenge lies in achieving accurate and reliable forecasts amidst fluctuating energy demand patterns and external factors such as weather and public events in the Caribbean context. This is one of the five ongoing projects at the Ilustre lab. ILUSTRE is a living lab in the Caribbean with the objective to develop, implement and test AI innovations that will accelerate the use of clean energy and advance solutions in water treatment and wastewater recycling/purification. The lab is a collaboration between JADS, the University of Curaçao, Lanubia Consulting, Aqualectra, WEB Bonaire, Alliander, KPN, CIWC, and the Ministry of Economic Development in Curaçao. Ilustre stands for Innovation Lab for Utilities on Sustainable Technology and Renewable Energy. It is one of the 17 Innovation Centers for Artificial Intelligence (ICAI) within the ROBUST Long-Term Programme (LTP). I also function as the lab manager of the Ilustre lab. (PhD candidate: Ugochukwi Orji)
The impact of climate on household poverty in Malawi and Nigeria, utilizing household panel surveys
The study addresses three key questions Projected Impact of Climate Variability: It aims to forecast the effects of climate variability on household poverty by 2050, using agent-based modeling and fuzzy-cognitive maps to integrate macro-level climate dynamics with micro-level decision-making processes. Characteristics of Affected Households: The study examines which households have been most affected by climate variability since 2010, employing the panel structure of the data and inferential statistics. It seeks to identify whether the impact of climate variability on poverty is consistent across different household types in Malawi and Nigeria. Effectiveness of Policy Interventions: The research simulates the distributional impact of various policy interventions, such as agricultural subsidies and social assistance programs, to assess their effectiveness at the household level. This analysis aims to inform evidence-based policy options to enhance household resilience in these countries. (PhD candidate: Hazal Colak Oz)
Resilient AI by Geometric Deep Learning
in this project we focus on knowledge graphs and applications of Geometric Deep Learning for Resilient AI.