Adam with averaging

This repository contains code related to the article Averaged Adam accelerates stochastic optimization in the training of deep neural network approximations for partial differential equation and optimal control problems by Steffen Dereich, Arnulf Jentzen and Adrian Riekert, which is currently available as a preprint on the arXiv.
The python scripts in the Scripts-directory implement the numerical examples from Section 3 in the paper.

Citation

[1] S. Dereich, A. Jentzen, A. Riekert.
Averaged Adam accelerates stochastic optimization in the training of deep neural network approximations for partial differential equation and optimal control problems.
Arxiv preprint arXiv:2501.06081, 2025.