How does PDE order affect the convergence of PINNs?

Authors: Chang hoon Song, Yesom Park, Myungjoo Kang

NeurIPS 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Furthermore, we present numerical experiments in support of our theoretical claims.
Researcher Affiliation Academia 1 Research Institute of Mathematics, Seoul National University 2 Department of Mathematics, University of California, Los Angeles 3 Department of Mathematical Sciences, Seoul National University
Pseudocode No The paper describes its methods mathematically and textually but does not include any clearly labeled 'Pseudocode' or 'Algorithm' blocks.
Open Source Code Yes Does the paper provide open access to the data and code, with sufficient instructions to faithfully reproduce the main experimental results, as described in supplemental material? Answer: [Yes]
Open Datasets No Throughout all experiments, the training collocation points consists of uniform grid and regularization parameters are set to ν1, . . . , νL = 1 and ν = 10. We implement all numerical experiments on a single NVIDIA RTX 3090 GPU.
Dataset Splits No The paper describes the number of 'training collocation points' and 'boundary points' used but does not explicitly detail dataset splits for validation or testing in the conventional machine learning sense (e.g., held-out portions of data for evaluation). The evaluation is done by comparing against the analytical solutions of the PDEs.
Hardware Specification Yes We implement all numerical experiments on a single NVIDIA RTX 3090 GPU.
Software Dependencies No Table 2 mentions 'optimizer(lr)' as 'GD(10-8)' or 'Adam(10-3)' but does not specify version numbers for these optimizers or any other software libraries or dependencies used in the experiments.
Experiment Setup Yes Throughout all experiments, the training collocation points consists of uniform grid and regularization parameters are set to ν1, . . . , νL = 1 and ν = 10. We trained networks with varying widths m, ranging from 102 to 106, for each combination of p and k using GD optimization with a learning rate of 10-8. Experimental details are provided in Appendix D. Table 2 provides experimental settings including optimizer, learning rate, and number of collocation points.