This repository contains code for the reproduction of some of the experiments in this paper with the same name.
The major improvement is the use of L-BFGS instead of Adam, which showed to work better in the "Physics-Informed Neural Network" setting.
The repo code for the reproductions of the paper mentioned
above is available under submission/extension_task. Whereas
the in the directory submission/base_task contains code
for slightly less difficult problems.
The file submission/report.pdf contains a description
of the report, mainly created as a submission for the course
Deep Learning in Scientific computing at ETH Zurich.