Stochastic Query Synthesis for Neural PDE Solvers

Daniel Musekamp, M.Sc; Mathias Niepert, Ph.D.

Ongoing Bachelor's Thesis

Abstract

PDEs are highly influential in physics and are describing various phenomena in the world, from wave movement to electro-magnetics. The problem arises when one tries to solve them, which requires enormous computing power for a numerical solution. Therefore, Musekamp et al. created a benchmark, which investigates active learning to get an approximation of the solution. In this work, we optimize the currently used pool-based sampling method by replacing it with stochastic query synthesis sampling. Instead of sampling from the pool set it will therefore be possible to sample directly from the input space, improving both diversity and efficiency.

Supervisors

To the top of the page