Scientific machine learning

Algebraic multigrid solvers in JAX.

A Julia library for building and training hybrid dynamic models that combine mechanistic and data-driven components.

A Julia package that breaks down the curse of dimensionality when solving nonlocal, nonlinear PDEs.

This project aims to develop a high-performance algebraic multigrid (AMG) solver in JAX that supports automatic differentiation and GPU acceleration, addressing critical gaps in existing Python and Julia implementations. The work will enable efficient gradient-based optimization in scientific machine learning applications requiring large-scale linear system solves, with potential to significantly impact the JAX ecosystem for differentiable physics simulations and inverse problems.

In this tutorial, you will learn about different techniques to infer parameters of a (differentiable) process-based model against data.