Can Synthetic Data Replace the Real Thing? An Information-Theoretic Look at Data Fidelity
Date:
Abstract
This talk explores evaluation of synthetic data fidelity across both tabular and time-series settings. Using generative models alongside rule-based simulations, this work examines how closely synthetic datasets reproduce real-world statistical, structural, and temporal patterns. The session provides an accessible, high-level overview of the methods, findings, and implications for the growing use of synthetic data in research, industry, and privacy-sensitive applications.
View Slides
]