Epistemic Intelligence & Computation Lab
College of Computing and Data Science, Nanyang Technological University, Singapore
The Epistemic Intelligence & Computation (EPIC) Lab is dedicated to advancing the principles necessary to developing epistemically intelligent systems, i.e. machine learning systems that can recognise, reason about, and communicate the limits of their knowledge. To achieve this, our research spans foundational theory to translational applications, focusing on knowledge-level uncertainty, commonly referred to as epistemic uncertainty.
-
Mathematical Foundations of Epistemic Uncertainty: We study the mathematical and statistical principles governing knowledge-level uncertainty, including representation (e.g. distributions v.s. sets), quantification (e.g. information-theoretic v.s. discrepency-based), comparisons (e.g. probability metrics), verification (e.g. statistical tests) of epistemic uncertainty. -
Algorithms for Epistemic Uncertainty-Aware Learning: We develope machine learning algorithms that can reason about, propagate, and communicate epistemic uncertainty in prediction (e.g. Gaussian process, neural networks, conformal framework), inference (e.g. causal inference), and decision-making (e.g. active learning, Bayesian optimisation). -
Epistemic Uncertainty-Aware AI Systems, Frameworks, and Applications: We translate our theoretical and algorithmic developments into practical AI systems within the broader trustworthy AI landscape, addressing challenges such as interpretability and explainability, out-of-domain detections, incentive-aware and adverserial machine learning,AI safety, robustness, and language models.
Why are we doing this? We believe that, although modern machine learning systems have achieved remarkable success in predictive and generative tasks — demonstrating powerful abilities to model statistical variability and patterns — their capacity to reason about uncertainty beyond observed data regularities remains fundamentally limited. In particular, forms of uncertainty arising from ignorance, ambiguity, and broader unknown-unknowns are not adequately captured by purely pattern-driven approaches. As reliance on such systems continues to grow, these limitations may lead to deeper structural challenges for reliability, robustness, and safety.
This gap motivates our research, which seeks to treat uncertainty as a first-class object of study rather than a secondary by-product of prediction. Our goal is to develop rigorous theory and algorithms to enable learning systems to represent, reason about, and communicate the limits of their knowledge — making uncertainty more than merely an afterthought or a set of error bars.
If any of this resonates with you, feel free to reach out for collaboration. We are always excited to hear from you.
The EPIC Lab is currently affiliated with the College of Computing and Data Science, Nanyang Technological University in Singapore. The lab maintains strong research ties with international collaborators across the United Kingdom, the Netherlands, Germany, Hong Kong, Japan, and Australia.
News
- Yusuf Sale (LMU Munich, Germany) visited our group and had an inspiring research discussion with us!
- Prof. Fabio Cuzzolin (Oxford Brookes University, UK) visited our group and delivered a research seminar at NTU!
- EPIC Lab coorganised the First workshop of Epistemic Intelligence in Machine Learning at EurIPS 2025!
Lab Gallery
