Accelerated Training of Physics-Informed Neural Networks (PINNs) using Meshless Discretizations

Authors: Ramansh Sharma, Varun Shankar

NeurIPS 2022 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We demonstrate the efficiency and accuracy of DT-PINNs via a series of experiments.
Researcher Affiliation Academia Ramansh Sharma Department of Computer Science and Engineering, SRM Institute of Science and Technology, India rs7146@srmist.edu.in Varun Shankar School of Computing, University of Utah, UT, USA shankar@cs.utah.edu
Pseudocode No No structured pseudocode or algorithm blocks were found in the paper.
Open Source Code Yes We release the datasets and codebase we used as part of the supplementary material.
Open Datasets No No specific link, DOI, repository name, or formal citation for a publicly available or open dataset was provided. The paper describes the generation of quasi-uniform collocation points and the use of manufactured solutions for its experiments.
Dataset Splits No The paper mentions 'training points' (collocation points) and a 'test set', but does not explicitly describe a validation set or the specific percentages/counts for training, validation, and test splits needed for reproduction.
Hardware Specification Yes All experiments were run for 5000 epochs on an NVIDIA Ge Force RTX 2070.
Software Dependencies No The paper mentions using PyTorch and CuPy in its references but does not provide specific version numbers for these or other software dependencies.
Experiment Setup Yes All experiments were run for 5000 epochs on an NVIDIA Ge Force RTX 2070. All results are reproducible with the seeds we used in the experiments. We used the L-BFGS optimizer with manually fine-tuned learning rates for both vanilla-PINNs and DT-PINNs. Both DT-PINNs and vanilla-PINNs used a constant NN depth of s = 4 layers with 50 nodes each across all runs.