Assistant Professor of Physics, UT Austin
William Gilpin studies nonlinear and high-dimensional systems using methods from physics, dynamical systems, statistical learning, and data analysis. The work includes fundamental questions about chaos and predictability, methods for reconstructing dynamics from time series, machine-learning models for physical systems, and applications in biophysics and quantitative biology.
Research on attractor reconstruction, predictability, recurrence, hidden drivers, and the mathematical structure of chaotic systems. This work uses chaos as both a physical phenomenon and a controlled setting for testing methods of statistical inference and forecasting.
Research on how neural networks and other statistical-learning methods represent, forecast, and generalize across dynamical systems. Topics include learned representations of physical rules, forecasting benchmarks, generative learning, pretrained time-series models, and transfer-operator interpretations of in-context learning.
Methods for extracting dynamical structure from partial or noisy observations. This work includes state-space reconstruction, recurrence analysis, latent-driver inference, forecasting, and the use of time-series foundation models for previously unseen systems.
Applications of nonlinear dynamics, statistical physics, and machine learning to biological systems, including cilia and biological fluid dynamics, systems biology, ecological transients, single-cell representations, and gene-network inference.
A chronological list of papers and preprints is available on the publications page.
Last updated August 2026.