Physics-informed neural networks (PINNs) have lately received great attention thanks to their flexibility in tackling a wide range of forward and inverse problems involving partial differential equations. However, despite their noticeable empirical success, little is known about how such constrained neural networks behave during their training via gradient descent. More importantly, even less is known about why such models sometimes fail to train at all. In this work, we aim to investigate these questions through the lens of the Neural Tangent Kernel (NTK); a kernel that captures the behavior of fully-connected neural networks in the infinite width limit during training via gradient descent. Specifically, we derive the NTK of PINNs and prove that, under appropriate conditions, it converges to a deterministic kernel that stays constant during training in the infinite-width limit. This allows us to analyze the training dynamics of PINNs through the lens of their limiting NTK and find a remarkable discrepancy in the convergence rate of the different loss components contributing to the total training error. To address this fundamental pathology, we propose a novel gradient descent algorithm that utilizes the eigenvalues of the NTK to adaptively calibrate the convergence rate of the total training error. Finally, we perform a series of numerical experiments to verify the correctness of our theory and the practical effectiveness of the proposed algorithms. The data and code accompanying this manuscript are publicly available at \url{https://github.com/PredictiveIntelligenceLab/PINNsNTK}.
Department of Computer Science, The Ohio State University（俄亥俄州立大学计算机科学系）；Halicioğlu Data Science Institute, University of California, San Diego（加利福尼亚大学圣迭戈分校哈利西奥卢数据科学学院）
The goal of this work is to shed light on the remarkable phenomenon of transition to linearity of certain neural networks as their width approaches infinity. We show that the transition to linearity of the model and, equivalently, constancy of the (neural) tangent kernel (NTK) result from the scaling properties of the norm of the Hessian matrix of the network as a function of the network width. We present a general framework for understanding the constancy of the tangent kernel via Hessian scaling applicable to the standard classes of neural networks. Our analysis provides a new perspective on…
Checkable NTK Positivity and Finite-Width Gradient Descent for Scalar- and Vector-Valued PINNs with Strong-Form, Weak-Form, and Nonlocal Linear Constraints
Is the neural tangent kernel of PINNs deep learning general partial differential equations always convergent ?
Academy of Mathematics and Systems Science, Chinese Academy of Sciences（中国科学院数学与系统科学研究院）；School of Mathematical Sciences, University of Chinese Academy of Sciences（中国科学院大学数学科学学院）
When and Why Adversarial Training Improves PINNs: A Neural Tangent Kernel Perspective
School of Mathematics and Statistics, Beijing Institute of Technology, China（北京理工大学数学与统计学院，中国）；School of Professional Education and Executive Development The Hong Kong Polytechnic University, China（香港理工大学专业教育学院及管理发展学院，中国）；Department of Computer Science & UCL AI Centre, University College London, UK（伦敦大学学院计算机科学系及UCL人工智能中心，英国）
Optimization dependent generalization bound for ReLU networks based on sensitivity in the tangent bundle
Institute for Computer Science and Control (SZTAKI)（计算机科学与控制研究所（SZTAKI））；Eötvös Loránd University（厄特沃什·罗兰大学）；CNRS CRIStAL（法国国家科学研究中心信号与自动化信息处理实验室）；Université de Lille（里尔大学）；Széchenyi István University（塞切尼·伊什特万大学）
Peking University（北京大学）；Georgia Institute of Technology（佐治亚理工学院）
Deep residual networks (ResNets) have demonstrated better generalization performance than deep feedforward networks (FFNets). However, the theory behind such a phenomenon is still largely unknown. This paper studies this fundamental problem in deep learning from a so-called "neural tangent kernel" perspective. Specifically, we first show that under proper conditions, as the width goes to infinity, training deep ResNets can be viewed as learning reproducing kernel functions with some kernel function. We then compare the kernel of deep ResNets with that of deep FFNets and discover that the class…
Unveiling the Mechanism of Continuous Representation Full-Waveform Inversion: A Wave Based Neural Tangent Kernel Framework
School of Mathematics and Statistics, Xi’an Jiaotong University, Xi’an, Shaanxi, China（西安交通大学数学与统计学学院）；Macao Institute of Systems Engineering, Macau University of Science and Technology, Taipa, Macao（澳门系统工程研究院，澳门科技大学，路环，澳门）
Ohio State University（俄亥俄州立大学）；UC, San Diego（加州大学圣地亚哥分校）
Wide neural networks with linear output layer have been shown to be near-linear, and to have near-constant neural tangent kernel (NTK), in a region containing the optimization path of gradient descent. These findings seem counter-intuitive since in general neural networks are highly complex models. Why does a linear structure emerge when the networks become wide? In this work, we provide a new perspective on this "transition to linearity" by considering a neural network as an assembly model recursively built from a set of sub-models corresponding to individual neurons. In this view, we show th…
Johns Hopkins University Applied Physics Laboratory（约翰斯·霍普金斯大学应用物理实验室）
In this work, we assess the ability of physics-informed neural networks (PINNs) to solve increasingly-complex coupled ordinary differential equations (ODEs). We focus on a pair of benchmarks: discretized partial differential equations and harmonic oscillators, each of which has a tunable parameter that controls its complexity. Even by varying network architecture and applying a state-of-the-art training method that accounts for "difficult" training regions, we show that PINNs eventually fail to produce correct solutions to these benchmarks as their complexity -- the number of equations and the…
Microsoft Research AI（微软研究院人工智能部门）；Stanford University（斯坦福大学）；University of Wisconsin-Madison（威斯康星大学麦迪逊分校）
The fundamental learning theory behind neural networks remains largely open. What classes of functions can neural networks actually learn? Why doesn't the trained network overfit when it is overparameterized? In this work, we prove that overparameterized neural networks can learn some notable concept classes, including two and three-layer networks with fewer parameters and smooth activations. Moreover, the learning can be simply done by SGD (stochastic gradient descent) or its variants in polynomial time using polynomially many samples. The sample complexity can also be almost independent of t…
An important inference from Neural Tangent Kernel (NTK) theory is the existence of spectral bias (SB), that is, low frequency components of the target function of a fully connected Artificial Neural Network (ANN) being learnt significantly faster than the higher frequencies during training. This is established for Mean Square Error (MSE) loss functions with very low learning rate parameters. Physics Informed Neural Networks (PINNs) are designed to learn the solutions of differential equations (DE) of arbitrary orders; in PINNs the loss functions are obtained as the residues of the conservative…
Physics-Informed Neural Networks have shown unique utility in parameterising the solution of a well-defined partial differential equation using automatic differentiation and residual losses. Though they provide theoretical guarantees of convergence, in practice the required training regimes tend to be exacting and demanding. Through the course of this paper, we take a deep dive into understanding the loss landscapes associated with a PINN and how that offers some insight as to why PINNs are fundamentally hard to optimise for. We demonstrate how PINNs can be forced to converge better towards th…
Establishing a theoretical analysis that explains why deep learning can outperform shallow learning such as kernel methods is one of the biggest issues in the deep learning literature. Towards answering this question, we evaluate excess risk of a deep learning estimator trained by a noisy gradient descent with ridge regularization on a mildly overparameterized neural network, and discuss its superiority to a class of linear estimators that includes neural tangent kernel approach, random feature model, other kernel methods, $k$-NN estimator and so on. We consider a teacher-student regression mo…
A fundamental bottleneck in utilising complex machine learning systems for critical applications has been not knowing why they do and what they do, thus preventing the development of any crucial safety protocols. To date, no method exist that can provide full insight into the granularity of the neural network's decision process. In the past, saliency maps were an early attempt at resolving this problem through sensitivity calculations, whereby dimensions of a data point are selected based on how sensitive the output of the system is to them. However, the success of saliency maps has been at be…
Deep neural networks can empirically perform efficient hierarchical learning, in which the layers learn useful representations of the data. However, how they make use of the intermediate representations are not explained by recent theories that relate them to "shallow learners" such as kernels. In this work, we demonstrate that intermediate neural representations add more flexibility to neural networks and can be advantageous over raw inputs. We consider a fixed, randomly initialized neural network as a representation function fed into another trainable network. When the trainable network is t…
Physics-informed neural networks (PINNs) are demonstrating remarkable promise in integrating physical models with gappy and noisy observational data, but they still struggle in cases where the target functions to be approximated exhibit high-frequency or multi-scale features. In this work we investigate this limitation through the lens of Neural Tangent Kernel (NTK) theory and elucidate how PINNs are biased towards learning functions along the dominant eigen-directions of their limiting NTK. Using this observation, we construct novel architectures that employ spatio-temporal and multi-scale ra…
Spectral Analysis of Hard-Constraint PINNs: The Spatial Modulation Mechanism of Boundary Functions
Sino--French Institute of Nuclear Engineering and Technology（中法核工程与技术研究院）；Sun Yat-sen University（中山大学）；CNPRI-SYSU Joint Research Center of Coolant Chemistry for Nuclear Reactor（中法核反应堆冷却剂化学联合研究中心）
IBM Research Europe（IBM欧洲研究院）；University of Modena and Reggio Emilia（摩德纳-雷焦艾米利亚大学）
A novel comparison is presented of the effect of optimiser choice on the accuracy of physics-informed neural networks (PINNs). To give insight into why some optimisers are better, a new approach is proposed that tracks the training trajectory curvature and can be evaluated on the fly at a low computational cost. The linear advection equation is studied for several advective velocities, and we show that the optimiser choice substantially impacts PINNs model performance and accuracy. Furthermore, using the curvature measure, we found a negative correlation between the convergence error and the c…
When PINNs Go Wrong: Pseudo-Time Stepping Against Spurious Solutions
Institute for Foundations of Data Science, Yale University（数据科学基础研究所，耶鲁大学）；Department of Mechanical Engineering and Applied Mechanics, University of Pennsylvania（机械工程与应用力学系，宾夕法尼亚大学）；Industrial Engineering & Management Science, Northwestern University（工业工程与管理科学，西北大学）
Physics-informed neural network (PINN) algorithms have shown promising results in solving a wide range of problems involving partial differential equations (PDEs). However, they often fail to converge to desirable solutions when the target function contains high-frequency features, due to a phenomenon known as spectral bias. In the present work, we exploit neural tangent kernels (NTKs) to investigate the training dynamics of PINNs evolving under stochastic gradient descent with momentum (SGDM). This demonstrates SGDM significantly reduces the effect of spectral bias. We have also examined why …
An efficient wavelet-based physics-informed neural network for multiscale problems
Department of Computational and Data Sciences, Indian Institute of Science, Bangalore, 560012, India（计算与数据科学系，印度科学研究院，班加罗尔，560012，印度）；Department of Mathematics, University of Manchester, Oxford Road, Manchester M13 9PL, UK（数学系，曼彻斯特大学，牛津路，曼彻斯特 M13 9PL，英国）
Technical University of Darmstadt（达姆施塔特工业大学）；RWTH Aachen University（亚琛工业大学）
Physics-informed neural networks (PINNs) are capable of finding the solution for a given boundary value problem. We employ several ideas from the finite element method (FEM) to enhance the performance of existing PINNs in engineering problems. The main contribution of the current work is to promote using the spatial gradient of the primary variable as an output from separated neural networks. Later on, the strong form which has a higher order of derivatives is applied to the spatial gradients of the primary variable as the physical constraint. In addition, the so-called energy form of the prob…
Why Rectified Power Unit Networks Fail and How to Improve It: An Effective Field Theory Perspective
School of Computational Sciences（计算科学学院）；Korea Institute for Advanced Study（韩国高级研究院）；Department of Mathematical Science（数学科学系）；Seoul National University（首尔国立大学）
Quantized neural networks have drawn a lot of attention as they reduce the space and computational complexity during the inference. Moreover, there has been folklore that quantization acts as an implicit regularizer and thus can improve the generalizability of neural networks, yet no existing work formalizes this interesting folklore. In this paper, we take the binary weights in a neural network as random variables under stochastic rounding, and study the distribution propagation over different layers in the neural network. We propose a quasi neural network to approximate the distribution prop…
Multi-dimensional training-priority weighting based on physical information propagation paths: a unified residual-weighting framework for physics-informed neural networks
Coupling-Robust Accuracy in Multiphysics Physics Informed Neural Networks via Kronecker-Preconditioned Optimization
Dept. of Control and Instrumentation Engineering, Korea University, Sejong, South Korea（控制与仪器工程系，韩国大学，世宗，韩国）；BK21 FOUR Smart Mobility Education and Research Team, Korea University, Sejong, South Korea（BK21 FOUR智能交通教育与研究团队，韩国大学，世宗，韩国）
