AI Foundations

AI Foundations develops the methods that enable intelligent systems to learn, represent information and make reliable predictions. The pillar brings together machine learning, generative modeling and computer vision.

Research directions

• Bayesian inference and uncertainty quantification.
• Generative models, information theory and stochastic processes.
• Representation learning, statistical learning and robustness.
• Computer vision and computer graphics.

From principles to reliable methods

We connect statistical and mathematical questions with the design of practical learning algorithms. Understanding uncertainty, data efficiency and robustness helps develop methods that remain useful when observations or computing resources are limited.

Connections across the unit

Foundational advances support medical learning and efficient intelligent systems, while challenges in these applications motivate new questions about inference, representation and learning.