Physics informed extreme learning machines with residual variation diminishing scheme for nonlinear problems with discontinuous surfaces

K Joshi, V Snigdha, AK Bhattacharya - IEEE Access, 2024 - ieeexplore.ieee.org
This work extends Extreme Learning Machines (ELM) to obtain solutions for nonlinear
higher order partial differential equations that govern the physics of different domains. The …

The Challenges of the Nonlinear Regime for Physics-Informed Neural Networks

A Bonfanti, G Bruno, C Cipriani - arXiv preprint arXiv:2402.03864, 2024 - arxiv.org
The Neural Tangent Kernel (NTK) viewpoint represents a valuable approach to examine the
training dynamics of Physics-Informed Neural Networks (PINNs) in the infinite width limit. We …

基于NTK 理论和改进时间因果的物理信息神经网络加速收敛算法

潘小果, 王凯, 邓维鑫 - 力学学报, 2024 - lxxb.cstam.org.cn
物理信息神经网络(physics-informed neural networks, PINNs) 是一类将先验物理知识嵌入神经
网络的方法, 目前已经成为求解偏微分方程领域的研究热点. 尽管PINNs 在数值模拟方面展现出 …

Physics-informed kernel learning

N Doumèche, F Bach, G Biau, C Boyer - arXiv preprint arXiv:2409.13786, 2024 - arxiv.org
Physics-informed machine learning typically integrates physical priors into the learning
process by minimizing a loss function that includes both a data-driven term and a partial …

Multiscale lubrication simulation based on fourier feature networks with trainable frequency

Y Tang, L Huang, L Wu, X Meng - arXiv preprint arXiv:2405.12638, 2024 - arxiv.org
Rough surface lubrication simulation is crucial for designing and optimizing tribological
performance. Despite the growing application of Physical Information Neural Networks …

Constructing Extreme Learning Machines with zero Spectral Bias

K Joshi, V Snigdha… - … Conference on Emerging …, 2023 - ieeexplore.ieee.org
The phenomena of Spectral Bias, where the higher frequency components of a function
being learnt in a feedforward Artificial Neural Network (ANN) are seen to converge more …

Data-Dependent Frequency Accumulated Physics-Informed Neural Networks for Steep Flow Field Prediction

P Xiaoguo, K Wang, D Weixin - Available at SSRN 4952223 - papers.ssrn.com
Physics-informed neural networks (PINNs) demonstrate remarkable ability to solve partial
differential equations (PDEs), while face challenges of high training cost and reaching high …