Adversarial Multi-task Learning Enhanced Physics-informed Neural Networks for Solving Partial Differential Equations

Recently, researchers have utilized neural networks to accurately solve partial differential equations (PDEs), enabling the mesh-free method for scientific computation. Unfortunately, the network performance drops when encountering a high nonlinearity domain. To improve the generalizability, we introduce the novel approach of employing multi-task learning techniques, the uncertainty-weighting loss and the gradients surgery, in the context of learning PDE solutions. The multi-task scheme exploits the benefits of learning shared representations, controlled by cross-stitch modules, between multiple related PDEs, which are obtainable by varying the PDE parameterization coefficients, to generalize better on the original PDE. Encouraging the network pay closer attention to the high nonlinearity domain regions that are more challenging to learn, we also propose adversarial training for generating supplementary high-loss samples, similarly distributed to the original training distribution. In the experiments, our proposed methods are found to be effective and reduce the error on the unseen data points as compared to the previous approaches in various PDE examples, including high-dimensional stochastic PDEs.

[1]  Roberto Cipolla,et al.  Multi-task Learning Using Uncertainty to Weigh Losses for Scene Geometry and Semantics , 2017, 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition.

[2]  Maziar Raissi,et al.  Forward-Backward Stochastic Neural Networks: Deep Learning of High-dimensional Partial Differential Equations , 2018, ArXiv.

[3]  Martial Hebert,et al.  Cross-Stitch Networks for Multi-task Learning , 2016, 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR).

[4]  Dimitrios I. Fotiadis,et al.  Artificial neural networks for solving ordinary and partial differential equations , 1997, IEEE Trans. Neural Networks.

[5]  C. Basdevant,et al.  Spectral and finite difference solutions of the Burgers equation , 1986 .

[6]  Vipin Kumar,et al.  Integrating Physics-Based Modeling with Machine Learning: A Survey , 2020, ArXiv.

[7]  Arnulf Jentzen,et al.  Solving high-dimensional partial differential equations using deep learning , 2017, Proceedings of the National Academy of Sciences.

[8]  Sebastian Ruder,et al.  Latent Multitask Architecture Learning , 2018 .

[9]  Andrew J. Davison,et al.  End-To-End Multi-Task Learning With Attention , 2018, 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR).

[10]  N. Phan-Thien,et al.  Neural-network-based approximations for solving partial differential equations , 1994 .

[11]  Jonathon Shlens,et al.  Explaining and Harnessing Adversarial Examples , 2014, ICLR.

[12]  Iasonas Kokkinos,et al.  UberNet: Training a Universal Convolutional Neural Network for Low-, Mid-, and High-Level Vision Using Diverse Datasets and Limited Memory , 2016, 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR).

[13]  Paris Perdikaris,et al.  Physics-informed neural networks: A deep learning framework for solving forward and inverse problems involving nonlinear partial differential equations , 2019, J. Comput. Phys..

[14]  Naftali Tishby,et al.  Machine learning and the physical sciences , 2019, Reviews of Modern Physics.

[15]  Anastasia Borovykh,et al.  Optimally weighted loss functions for solving PDEs with Neural Networks , 2020, ArXiv.

[16]  Justin A. Sirignano,et al.  DGM: A deep learning algorithm for solving partial differential equations , 2017, J. Comput. Phys..

[17]  Jimmy Ba,et al.  Adam: A Method for Stochastic Optimization , 2014, ICLR.

[18]  Zhao Chen,et al.  GradNorm: Gradient Normalization for Adaptive Loss Balancing in Deep Multitask Networks , 2017, ICML.

[19]  Joan Bruna,et al.  Intriguing properties of neural networks , 2013, ICLR.

[20]  S. Levine,et al.  Gradient Surgery for Multi-Task Learning , 2020, NeurIPS.

[21]  Paris Perdikaris,et al.  Adversarial Uncertainty Quantification in Physics-Informed Neural Networks , 2018, J. Comput. Phys..

[22]  Tim Dockhorn,et al.  A Discussion on Solving Partial Differential Equations using Neural Networks , 2019, ArXiv.

[23]  Yoshua Bengio,et al.  Generative Adversarial Nets , 2014, NIPS.

[24]  M. Stein Large sample properties of simulations using latin hypercube sampling , 1987 .

[25]  Ameet Talwalkar,et al.  Massively Parallel Hyperparameter Tuning , 2018, ArXiv.

[26]  Ali Al-Aradi,et al.  Solving Nonlinear and High-Dimensional Partial Differential Equations via Deep Learning , 2018, 1811.08782.

[27]  Masayuki Numao,et al.  Physics-Guided Neural Network with Model Discrepancy Based on Upper Troposphere Wind Prediction , 2019, 2019 18th IEEE International Conference On Machine Learning And Applications (ICMLA).

[28]  Jasper Snoek,et al.  Practical Bayesian Optimization of Machine Learning Algorithms , 2012, NIPS.

[29]  Barak A. Pearlmutter,et al.  Automatic differentiation in machine learning: a survey , 2015, J. Mach. Learn. Res..

[30]  Panos Parpas,et al.  Towards Robust and Stable Deep Learning Algorithms for Forward Backward Stochastic Differential Equations , 2019, ArXiv.