Skip to content

Latest commit

 

History

History
24 lines (18 loc) · 1.03 KB

README.md

File metadata and controls

24 lines (18 loc) · 1.03 KB

Physical Modelling of Stiff Membrane Vibration Using Neural Networks with Spectral Convolution Layers

Code for the paper Physical Modelling of Stiff Membrane Vibration Using Neural Networks with Spectral Convolution Layers

Reproducing the results

GPU needed, might need to adjust batch size depending on GPU RAM

Create the environment

conda env create -f environment.yml

test run

conda activate platefno && python train_2d_plate.py -m +experiment=testrun solver/damping=freqdependent

Train the models

conda activate platefno && python train_2d_plate.py -m +experiment=pdeparamsweep solver/damping=freqdependent

get validation results and generate tex table

conda activate platefno && python validate_2d_plate.py ./output/pdeparamsweep

get plots for the extrapolation

The highlighted runs shown in the paper are gamma = 1.0, kappa = 0.1 run 4 gamma = 1.0, kappa = 1.0 run 4 gamma = 100.0, kappa = 0.1 run 2 conda activate platefno && python evaluate_extrapolation.py ./output/pdeparamsweep