Ludger Paehler

Scalable Automatic Differentiation of Multiple Parallel Paradigms through Compiler Augmentation

Composable automatic differentiation of parallel programming paradigms at the compiler level. Published at SC'22.

On the Relationships between Graph Neural Networks for the Simulation of Physical Systems and Classical Numerical Methods

An overview on the parallels between the development of numerical approaches to the development of machine learning models to make machine learning models for science more efficient. Published at the AI4Science workshop at ICML'22.

Reverse-Mode Automatic Differentiation and Optimization of GPU Kernels via Enzyme

Syndication of the publication on the first general automatic differentiation of CUDA & ROCm kernels on the LLVM-level. Published at the Beyond Bayes workshop at ICML'22.

Reverse-Mode Automatic Differentiation and Optimization of GPU Kernels via Enzyme

First general automatic differentiation of CUDA & ROCm kernels on the LLVM-level. Published at SC'21.

Sparse Identification of Truncation Errors

Applying sparse regression to the modified differential equation to identify the truncation errors of numerical schemes. Published in JCP'19.