Relu Deep Neural Networks and Linear Finite Elements

In this paper, we investigate the relationship between deep neural networks (DNN) with rectified linear unit (ReLU) function as the activation function and continuous piecewise linear (CPWL) functions, especially CPWL functions from the simplicial linear finite element method (FEM). We first consider the special case of FEM. By exploring the DNN representation of its nodal basis functions, we present a ReLU DNN representation of CPWL in FEM. We theoretically establish that at least $2$ hidden layers are needed in a ReLU DNN to represent any linear finite element functions in $\Omega \subseteq \mathbb{R}^d$ when $d\ge2$. Consequently, for $d=2,3$ which are often encountered in scientific and engineering computing, the minimal number of two hidden layers are necessary and sufficient for any CPWL function to be represented by a ReLU DNN. Then we include a detailed account on how a general CPWL in $\mathbb R^d$ can be represented by a ReLU DNN with at most $\lceil\log_2(d+1)\rceil$ hidden layers and we also give an estimation of the number of neurons in DNN that are needed in such a representation. Furthermore, using the relationship between DNN and FEM, we theoretically argue that a special class of DNN models with low bit-width are still expected to have an adequate representation power in applications. Finally, as a proof of concept, we present some numerical results for using ReLU DNNs to solve a two point boundary problem to demonstrate the potential of applying DNN for numerical solution of partial differential equations.

[1]  Alexander Cloninger,et al.  Provable approximation properties for deep neural networks , 2015, ArXiv.

[2]  Allan Pinkus,et al.  Approximation theory of the MLP model in neural networks , 1999, Acta Numerica.

[3]  Pingwen Zhang,et al.  Moving mesh methods in multiple dimensions based on harmonic maps , 2001 .

[4]  E Weinan,et al.  The Deep Ritz Method: A Deep Learning-Based Numerical Algorithm for Solving Variational Problems , 2017, Communications in Mathematics and Statistics.

[5]  Bin Liu,et al.  Ternary Weight Networks , 2016, ICASSP 2023 - 2023 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP).

[6]  Yoshua Bengio,et al.  Gradient-based learning applied to document recognition , 1998, Proc. IEEE.

[7]  Marcello Sanguineti,et al.  Comparison of worst case errors in linear and neural network approximation , 2002, IEEE Trans. Inf. Theory.

[8]  Hrushikesh Narhar Mhaskar,et al.  On the tractability of multivariate integration and approximation by neural networks , 2004, J. Complex..

[9]  Shuning Wang,et al.  Generalization of hinging hyperplanes , 2005, IEEE Transactions on Information Theory.

[10]  Kurt Hornik,et al.  Multilayer feedforward networks are universal approximators , 1989, Neural Networks.

[11]  L. Jones.  A Simple Lemma on Greedy Approximation in Hilbert Space and Convergence Rates for Projection Pursuit Regression and Neural Network Training , 1992 .

[12]  Jack Xin,et al.  BinaryRelax: A Relaxation Approach For Training Deep Neural Networks With Quantized Weights , 2018, SIAM J. Imaging Sci..

[13]  Raman Arora,et al.  Understanding Deep Neural Networks with Rectified Linear Units , 2016, Electron. Colloquium Comput. Complex..

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

[15]  E Weinan,et al.  Deep Learning-Based Numerical Methods for High-Dimensional Parabolic Partial Differential Equations and Backward Stochastic Differential Equations , 2017, Communications in Mathematics and Statistics.

[16]  Andrew R. Barron,et al.  Universal approximation bounds for superpositions of a sigmoidal function , 1993, IEEE Trans. Inf. Theory.

[17]  Lexing Ying,et al.  Solving parametric PDE problems with artificial neural networks , 2017, European Journal of Applied Mathematics.

[18]  Ding-Xuan Zhou,et al.  Universality of Deep Convolutional Neural Networks , 2018, Applied and Computational Harmonic Analysis.

[19]  Nadav Cohen,et al.  On the Expressive Power of Deep Learning: A Tensor Analysis , 2015, COLT 2016.

[20]  Allan Pinkus,et al.  Multilayer Feedforward Networks with a Non-Polynomial Activation Function Can Approximate Any Function , 1991, Neural Networks.

[21]  S. W. Ellacott,et al.  Aspects of the numerical analysis of neural networks , 1994, Acta Numerica.