IDEAS develops and applies data-driven methods to urgent challenges in environmental and life sciences—often involving large, heterogeneous datasets and high uncertainty.
Two application pillars
- Life Sciences & Health — From biological data science to health-related applications, IDEAS projects address complex systems where data-driven models can accelerate discovery and decision-making.
- Environmental Sciences — IDEAS tackles questions shaped by spatio-temporal dynamics, changing conditions, and non-i.i.d. data—where robustness and credible inference are critical.
Research topics
- Hybrid Learning & Physics-Based AI — Integrating process-based models with ML.
- Unstructured Data Mining & Harmonization — Using text, images, omics, citizen science.
- Explainability & Credibility — Interpretable, trustworthy models; links to causal inference.
- Robustness in Non-I.I.D. Data — Reliable learning under dataset shift and spatio-temporal correlation.
- Generative AI / Foundation Models with Domain Knowledge — Generative models grounded in domain expertise.
Cross-cutting priorities
IDEAS emphasizes uncertainty quantification and improved digital processes in research data management to produce AI-ready FAIR data—essential for trust in scientific predictions.
