1 paper
Nilo Schwencke, Cyriaque Rousselot, Alena Shilova +1
Recent works have shown that natural gradient methods can significantly outperform standard optimizers when training physics-informed neural networks (PINNs). In this paper, we ana…