Operator Learning on viscous Burgers' equation n=8192
0.001Relative Error (x10^-3)Galerkin Transformer
Evaluation Results
| Method | Links | |
|---|---|---|
| Galerkin TransformerNormalization scheme=Galerkin-type LayerNorm on K, V2021.05 | 0.001 | |
| Fourier TransformerNormalization scheme=Galerkin-type LayerNorm on Q, K2021.05 | 0.0011 | |
| Fourier TransformerNormalization scheme=Regular LayerNorm2021.05 | 0.0012 | |
| Linear TransformerNormalization scheme=Galerkin-type LayerNorm on K, V2021.05 | 0.0012 | |
| Softmax TransformerNormalization scheme=Galerkin-type LayerNorm on Q, K2021.05 | 0.0013 | |
| Linear TransformerNormalization scheme=Regular LayerNorm2021.05 | 0.0016 | |
| Softmax TransformerNormalization scheme=Regular LayerNorm2021.05 | 0.002 | |
| Galerkin TransformerNormalization scheme=Regular LayerNorm2021.05 | 0.0027 | |
| FNO1dScheduler=1cycle, Epochs=1002021.05 | 0.0042 | |
| FNO1dScheduler=Step, Epochs=5002021.05 | 0.0139 |