Learning to Predict Motion from a Few Examples
Professor: Siting Liu
Description: My group develops in-context learning methods that use a few examples to predict the behavior of a new system without retraining the model. The Cal-Bridge scholar will explore this idea using spring–mass systems, testing whether example trajectories allow a trained model to predict motion from a new starting position or velocity. With faculty-provided simulation and machine-learning code, the scholar will generate data, visualize predictions, and investigate how prediction accuracy depends on the number of examples and changes in spring stiffness or damping.
Preferred Qualifications: Basic programming experience, preferably in Python, and introductory calculus. No prior experience with machine learning or differential equations is required; relevant concepts will be introduced during the project.
