Notice: The reproducibility variables underlying each score are classified using an automated LLM-based pipeline, validated against a manually labeled dataset. LLM-based classification introduces uncertainty; scores should be interpreted as estimates. Full accuracy metrics and methodology are described in [1]

Learning to Accelerate Partial Differential Equations via Latent Global Evolution

Authors: Tailin Wu, Takashi Maruyama, Jure Leskovec

NeurIPS 2022 | Venue PDF | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We test our method in a 1D benchmark of nonlinear PDEs, 2D Navier-Stokes flows into turbulent phase and an inverse optimization of boundary conditions in 2D Navier-Stokes flow.
Researcher Affiliation Collaboration Tailin Wu Department of Computer Science Stanford University EMAIL Takashi Maruyama NEC Corp. & Stanford University EMAIL & EMAIL Jure Leskovec Department of Computer Science Stanford University EMAIL
Pseudocode No The paper does not contain a pseudocode block or an algorithm block.
Open Source Code Yes 1Project website and code can be found at http://snap.stanford.edu/le_pde/.
Open Datasets Yes We use the 1D benchmark in [7], whose PDEs are... We test LE-PDE in a 2D benchmark [14] of Navier-Stokes equation.
Dataset Splits No We perform hyperparameter search over latent dimension of {64, 128} and use the model with best validation performance. While this implies a validation set was used, no specific details about the data split (e.g., percentages or counts) are provided in the main text.
Hardware Specification No The paper states "Did you include the total amount of compute and the type of resources used (e.g., type of GPUs, internal cluster, or cloud provider)? [Yes] In Appendix D,E,F,G." This indicates hardware details are in the appendix, but they are not explicitly described in the main text itself.
Software Dependencies No The paper mentions "PhiFlow [73] as our ground-truth solver", but does not specify a version number for this or any other key software dependencies.
Experiment Setup Yes We perform hyperparameter search over latent dimension of {64, 128} and use the model with best validation performance. To ensure a fair comparison, here our LE-PDE uses past 10 steps to predict one future step and temporal bundling S = 1 (no bundling), the same setting as in FNO-2D.