Dataset and Model Weights for Plasma Sheet Model Graph Network Simulator

This repository contains the simulation data and pre-trained Graph Neural Network (GNN) models produced in [1].

Two *.zip files are provided:

• data.zip - contains the datasets of train/test simulations produced using the Sheet Model algorithm [1, 2]
• models.zip - contains the GNN model weights (*.pkl) + relevant training information and model parameters (*.yml and *.txt)

Dataset subfolders are named according to dataset/{'train' or 'test'}/{number of sheets}/{boundary condition}/. Each subfolder contains multiple simulations and a single info.yml file with relevant information regarding the overall setup. For each i-th simulation the following files are provided:

• x_{i}.npy - array with sheet trajectories (#time-steps, #sheets)
• v_{i}.npy - array with sheet velocities (#time-steps, #sheets)
• x_eq_{i}.npy - array with sheet equilibrium positions (#time-steps, #sheets)

 Model sub-folders are named according to :

• models/{time step}/{seed} - default architecture (preferred)
• models/{time step}/{'collisions', 'nosent' or 'equivariant'}/{seed} - alternative (less performing) architectures mentioned in the paper appendices.

For each model we provide:

• params_best.pkl - model weights that performed the best during training on the validation set
• params_final.pkl - model weights at the end of training
• model_cfg.yml - GNN architecture metadata
• train_cfg.yml - training configuration metadata
• train_data.yml - training dataset metadata
• loss.txt - training and validation loss per epoch
• loss_i.txt - training loss per gradient update step

Source Code

The source code used to produce the data, train, and test the models can be found at: https://github.com/diogodcarvalho/gns-sheet-model

References

[1] D. D. Carvalho, D. R. Ferreira, L. O. Silva, "Learning the dynamics of a one-dimensional plasma model with graph neural networks", Mach. Learn.: Sci. Technol. 5 025048 (2024)

[2] J. Dawson, "One‐Dimensional Plasma Model", The Physics of Fluids 5.4 (1962): 445-459.