C1 – An Introduction to Generative Models

Chair: Yuan Wu, PhD (Duke University)

Instructor: Anru Zhang, PhD (Duke University)

Course Description:

Generative modeling is a rapidly evolving area in data science that enables the creation, analysis, and interpretation of complex data. This short course introduces the statistical foundations of generative models and their growing impact on biomedical research.

We will begin with backgrounds, variational autoencoders, generative adversarial networks, and normalizing flows, before moving to state-of-the-art approaches such as diffusion models and flow matching. Throughout, we will emphasize how these models relate to fundamental statistical ideas, including estimation, uncertainty quantification, and bias correction.

Applications will focus on biomedical science, showcasing how generative methods enable protein structure prediction, design, and simulation. Case studies will demonstrate how these approaches address practical challenges such as small sample sizes, irregular time series, and sampling bias, enabling data augmentation and accelerating discovery.

The course is designed for a broad audience across statistics, machine learning, and the biomedical sciences. By integrating theory, methodology, and applications, it aims to equip participants with a principled understanding of generative modeling and inspire future research.