Track papers across labs, topics, and trends.
Search papers, labs, and topics across Lattice.
This paper introduces "ghost tasking," a novel procedure for parametrized Gaussian Processes (GPs) to solve linear differential equations, particularly in low-data regimes where existing methods struggle with non-parametrizable systems or excessive task requirements. By constructing auxiliary "ghost" tasks, the method effectively renders any system parametrizable, enabling algorithmic GP construction with minimal latent functions and output dimensions. Ghost tasking demonstrates superior performance, especially in inverse problems with scarce data, outperforming existing methods in empirical evaluations.
Auxiliary "ghost tasks" make any linear differential equation system effectively parametrizable for Gaussian Processes, solving a key limitation in physics-informed ML.
Physics-informed machine learning has gained significant attention in recent years. In regimes of limited data, parametrized Gaussian processes have become popular. Existing approaches, however, often face limitations, such as requiring parametrizable (also called controllable) systems or a large number of output tasks. In this work, we introduce a systematic procedure we call"ghost tasking", using auxiliary tasks to circumvent these limitations. We prove that such ghost tasks can render any non-parametrizable system effectively parametrizable, enabling algorithmic construction of parametrized Gaussian Processes while keeping the number of required tasks (i.e. output dimensions) and latent functions low. We find that ghost tasking performs especially well in an inverse problem setting, even with very few available data. We show the usage and power of ghost tasking in three experiments, providing systematic comparisons to the only other currently available method applicable to all experiments. We provide necessary syntax and explications for two computer algebra programs that compute parametrizations for systems with polynomial or rational coefficients. Our theoretical results extend to systems with meromorphic functions.