Skip to content

Benchmarks

This is an overview of some benchmark results to compare the performance of different operator architectures on various problems.

The benchmarks are implemented in the benchmarks directory and we refer to this directory for detailed information on how the benchmarks are run.

Reference: Li, Zongyi, et al. "Fourier neural operator for parametric partial differential equations." arXiv preprint arXiv:2010.08895 (2020) Table 1 (\(\nu\) = 1e−5 T=20 N=1000)

reported for FNO-3D: 0.1893 (rel. test error)

FourierNeuralOperator

Depth Parameters Training Time (V100) rel. train error rel. test error
4 201M 91min 4.94e-03 0.1862
16 805M 181min 1.84e-04 0.1486
Visualization of best and worst training and test samples for FNO with depth 4 (as in paper).
Best training sample
Best training sample rel. error = 2.3351e-03 		Worst training sample
Worst training sample rel. error = 1.0090e-02 		Best test sample
Best test sample rel. error = 1.0219e-01 		Worst test sample
Worst test sample rel. error = 4.4294e-01
Last update: 2024-08-20
Created: 2024-08-20