American Institute of Mathematical Sciences

Advanced
August  2020, 19(8): 4085-4095. doi: 10.3934/cpaa.2020181

Function approximation by deep networks

H. N. Mhaskar 1,, and  T. Poggio 2,

1. 

Institute of Mathematical Sciences, Claremont Graduate University, Claremont, CA 91711
2. 

Center for Brains, Minds, and Machines, McGovern Institute for Brain Research, Massachusetts Institute of Technology, Cambridge, MA, 02139

*Corresponding author

Received  August 2019 Revised  November 2019 Published  May 2020

Fund Project: The research of the first author is supported in part by the Office of the Director of National Intelligence (ODNI), Intelligence Advanced Research Projects Activity (IARPA), via 2018-18032000002. The research of the second author is supported by the Center for Brains, Minds and Machines (CBMM), funded by NSF STC award CCF-1231216

We show that deep networks are better than shallow networks at approximating functions that can be expressed as a composition of functions described by a directed acyclic graph, because the deep networks can be designed to have the same compositional structure, while a shallow network cannot exploit this knowledge. Thus, the blessing of compositionality mitigates the curse of dimensionality. On the other hand, a theorem called good propagation of errors allows to "lift" theorems about shallow networks to those about deep networks with an appropriate choice of norms, smoothness, etc. We illustrate this in three contexts where each channel in the deep network calculates a spherical polynomial, a non-smooth ReLU network, or another zonal function network related closely with the ReLU network.

Keywords: Deep networks, degree of approximation, approximation on the Euclidean sphere.
Mathematics Subject Classification: Primary: 41A25; Secondary: 68Q32.
Citation: H. N. Mhaskar, T. Poggio. Function approximation by deep networks. Communications on Pure & Applied Analysis, 2020, 19 (8) : 4085-4095. doi: 10.3934/cpaa.2020181
References:
[1]

F. Bach, Breaking the curse of dimensionality with convex neural networks, J. Mach. Learn. Res., 18 (2017), 629-681.   Google Scholar

[2]

Y. Cho and L. K. Saul, Kernel methods for deep learning, in Advances in Neural Information Processing Systems, (2009), 342–350. Google Scholar

[3]

C. K. Chui, X. Li and H. N. Mhaskar, Limitations of the approximation capabilities of neural networks with one hidden layer, Adv. Comput. Math., 5 (1996), 233-243.  doi: 10.1007/BF02124745.  Google Scholar

[4]

C. K. Chui, S. B. Lin and D. X. Zhou, Construction of neural networks for realization of localized deep learning, Front. Appl. Math. Statist., 4 (2018). doi: 10.1109/tnnls.2017.2665555.  Google Scholar

[5]

R. Eldan and O. Shamir, The power of depth for feedforward neural networks, in Conference on Learning Theory, (2016), 907–940. Google Scholar

[6]

B. Hanin, Universal function approximation by deep neural nets with bounded width and relu activations, Mathematics, 7 (2019), Art. 992. doi: 10.3390/math7100992.  Google Scholar

[7]

Q. T. Le Gia and H. N. Mhaskar, Localized linear polynomial operators and quadrature formulas on the sphere, SIAM J. Numer. Anal., 47 (2009), 440-466.  doi: 10.1137/060678555.  Google Scholar

[8]

P. Lizorkin and K. P. Rustamov, Nikol'skii-Besov spaces on the sphere in connection with approximation theory, Proc. Steklov Inst. Math. AMS Trans., 204 (1994), 149-172.   Google Scholar

[9]

H. N. Mhaskar, Approximation properties of a multilayered feedforward artificial neural network, Adv. Comput. Math., 1 (1993), 61-80.  doi: 10.1007/BF02070821.  Google Scholar

[10]

H. N. Mhaskar, Eignets for function approximation on manifolds, Appl. Comput. Harmon. Anal., 29 (2010), 63-87.  doi: 10.1016/j.acha.2009.08.006.  Google Scholar

[11]

H. N. Mhaskar, Dimension independent bounds for general shallow networks, Neural Netw., 123 (2020), 142-152.  doi: 10.1016/j.neunet.2019.11.006.  Google Scholar

[12]

H. N. Mhaskar, Function approximation with zonal function networks with activation functions analogous to the rectified linear unit functions, J. Complexity, 51 (2019), 1-19.  doi: 10.1016/j.jco.2018.09.002.  Google Scholar

[13]

H. N. Mhaskar and T. Poggio, Deep vs. shallow networks: An approximation theory perspective, Anal. Appl., 14 (2016), 829-848.  doi: 10.1142/S0219530516400042.  Google Scholar

[14]

R. Montufar, G. F., Pa scanu, K. Cho and Y. Bengio, On the number of linear regions of deep neural networks, Adv. Neural Inform. Process. Syst., 27 (2014), 2924-2932.   Google Scholar

[15]

S. Pawelke, Über die Approximationsordnung bei Kugelfunktionen und algebraischen Polynomen, Tohoku Math. J. Sec. Ser., 24 (1972), 473-486.  doi: 10.2748/tmj/1178241489.  Google Scholar

[16]

I. Safran and O. Shamir, Depth separation in relu networks for approximating smooth non-linear functions, preprint, arXiv: 1610.09887. Google Scholar

[17]

I. Safran and O. Shamir, Depth-width tradeoffs in approximating natural functions with neural networks, in Proceedings of the 34th International Conference on Machine Learning, Vol. 70, (2017), 2979–2987. Google Scholar

[18]

T. Serra, C. Tjandraatmadja and S. Ramalingam, Bounding and counting linear regions of deep neural networks, preprint, arXiv: 1711.02114. Google Scholar

[19]

O. Sharir and A. Shashua, On the expressive power of overlapping architectures of deep learning, preprint, arXiv: 1703.02065. Google Scholar

[20]

M. Telgarsky, Benefits of depth in neural networks, preprint, arXiv: 1602.04485. Google Scholar

[21]

D. Yarotsky, Error bounds for approximations with deep relu networks, Neural Netw., 94 (2017), 103-114.   Google Scholar

[22]

D. Yarotsky, Optimal approximation of continuous functions by very deep relu networks, preprint, arXiv: 1802.03620. Google Scholar

show all references

References:
[1]

F. Bach, Breaking the curse of dimensionality with convex neural networks, J. Mach. Learn. Res., 18 (2017), 629-681.   Google Scholar

[2]

Y. Cho and L. K. Saul, Kernel methods for deep learning, in Advances in Neural Information Processing Systems, (2009), 342–350. Google Scholar

[3]

C. K. Chui, X. Li and H. N. Mhaskar, Limitations of the approximation capabilities of neural networks with one hidden layer, Adv. Comput. Math., 5 (1996), 233-243.  doi: 10.1007/BF02124745.  Google Scholar

[4]

C. K. Chui, S. B. Lin and D. X. Zhou, Construction of neural networks for realization of localized deep learning, Front. Appl. Math. Statist., 4 (2018). doi: 10.1109/tnnls.2017.2665555.  Google Scholar

[5]

R. Eldan and O. Shamir, The power of depth for feedforward neural networks, in Conference on Learning Theory, (2016), 907–940. Google Scholar

[6]

B. Hanin, Universal function approximation by deep neural nets with bounded width and relu activations, Mathematics, 7 (2019), Art. 992. doi: 10.3390/math7100992.  Google Scholar

[7]

Q. T. Le Gia and H. N. Mhaskar, Localized linear polynomial operators and quadrature formulas on the sphere, SIAM J. Numer. Anal., 47 (2009), 440-466.  doi: 10.1137/060678555.  Google Scholar

[8]

P. Lizorkin and K. P. Rustamov, Nikol'skii-Besov spaces on the sphere in connection with approximation theory, Proc. Steklov Inst. Math. AMS Trans., 204 (1994), 149-172.   Google Scholar

[9]

H. N. Mhaskar, Approximation properties of a multilayered feedforward artificial neural network, Adv. Comput. Math., 1 (1993), 61-80.  doi: 10.1007/BF02070821.  Google Scholar

[10]

H. N. Mhaskar, Eignets for function approximation on manifolds, Appl. Comput. Harmon. Anal., 29 (2010), 63-87.  doi: 10.1016/j.acha.2009.08.006.  Google Scholar

[11]

H. N. Mhaskar, Dimension independent bounds for general shallow networks, Neural Netw., 123 (2020), 142-152.  doi: 10.1016/j.neunet.2019.11.006.  Google Scholar

[12]

H. N. Mhaskar, Function approximation with zonal function networks with activation functions analogous to the rectified linear unit functions, J. Complexity, 51 (2019), 1-19.  doi: 10.1016/j.jco.2018.09.002.  Google Scholar

[13]

H. N. Mhaskar and T. Poggio, Deep vs. shallow networks: An approximation theory perspective, Anal. Appl., 14 (2016), 829-848.  doi: 10.1142/S0219530516400042.  Google Scholar

[14]

R. Montufar, G. F., Pa scanu, K. Cho and Y. Bengio, On the number of linear regions of deep neural networks, Adv. Neural Inform. Process. Syst., 27 (2014), 2924-2932.   Google Scholar

[15]

S. Pawelke, Über die Approximationsordnung bei Kugelfunktionen und algebraischen Polynomen, Tohoku Math. J. Sec. Ser., 24 (1972), 473-486.  doi: 10.2748/tmj/1178241489.  Google Scholar

[16]

I. Safran and O. Shamir, Depth separation in relu networks for approximating smooth non-linear functions, preprint, arXiv: 1610.09887. Google Scholar

[17]

I. Safran and O. Shamir, Depth-width tradeoffs in approximating natural functions with neural networks, in Proceedings of the 34th International Conference on Machine Learning, Vol. 70, (2017), 2979–2987. Google Scholar

[18]

T. Serra, C. Tjandraatmadja and S. Ramalingam, Bounding and counting linear regions of deep neural networks, preprint, arXiv: 1711.02114. Google Scholar

[19]

O. Sharir and A. Shashua, On the expressive power of overlapping architectures of deep learning, preprint, arXiv: 1703.02065. Google Scholar

[20]

M. Telgarsky, Benefits of depth in neural networks, preprint, arXiv: 1602.04485. Google Scholar

[21]

D. Yarotsky, Error bounds for approximations with deep relu networks, Neural Netw., 94 (2017), 103-114.   Google Scholar

[22]

D. Yarotsky, Optimal approximation of continuous functions by very deep relu networks, preprint, arXiv: 1802.03620. Google Scholar

13] shows an example of a $ \mathcal{G} $–function ($ f^* $ given in (3.1)). The vertices $ V\cup \mathbf{S} $ of the DAG $ \mathcal{G} $ are denoted by red dots. The black dots represent the inputs; the input to the various nodes as indicated by the in–edges of the red nodes. The blue dot indicates the output value of the $ \mathcal{G} $–function, $ f^* $ in this example">Figure 1.  This figure from [13] shows an example of a $ \mathcal{G} $–function ($ f^* $ given in (3.1)). The vertices $ V\cup \mathbf{S} $ of the DAG $ \mathcal{G} $ are denoted by red dots. The black dots represent the inputs; the input to the various nodes as indicated by the in–edges of the red nodes. The blue dot indicates the output value of the $ \mathcal{G} $–function, $ f^* $ in this example
Figure Options

Download full-size image

Download as PowerPoint slide

Figure 2.  On the left, with $ {\mathbf{x}}_0 = (1, 1, 1)/\sqrt{3} $, the graph of $ f({\mathbf{x}}) = [({\mathbf{x}}\cdot{\mathbf{x}}_0-0.1)_+]^8 + [(-{\mathbf{x}}\cdot{\mathbf{x}}_0-0.1)_+]^8 $. On the right, the graph of $ \mathcal{D}_{\phi_\gamma}(f) $. Courtesy: D. Batenkov
Figure Options

Download full-size image

Download as PowerPoint slide

[1]

Weihua Liu, Andrew Klapper. AFSRs synthesis with the extended Euclidean rational approximation algorithm. Advances in Mathematics of Communications, 2017, 11 (1) : 139-150. doi: 10.3934/amc.2017008
[2]

Vikas S. Krishnamurthy. The vorticity equation on a rotating sphere and the shallow fluid approximation. Discrete & Continuous Dynamical Systems, 2019, 39 (11) : 6261-6276. doi: 10.3934/dcds.2019273
[3]

Purshottam Narain Agrawal, Şule Yüksel Güngör, Abhishek Kumar. Better degree of approximation by modified Bernstein-Durrmeyer type operators. Mathematical Foundations of Computing, 2021  doi: 10.3934/mfc.2021024
[4]

Gabriella Bretti, Roberto Natalini, Benedetto Piccoli. Fast algorithms for the approximation of a traffic flow model on networks. Discrete & Continuous Dynamical Systems - B, 2006, 6 (3) : 427-448. doi: 10.3934/dcdsb.2006.6.427
[5]

Pierluigi Colli, Gianni Gilardi, Jürgen Sprekels. Deep quench approximation and optimal control of general Cahn–Hilliard systems with fractional operators and double obstacle potentials. Discrete & Continuous Dynamical Systems - S, 2021, 14 (1) : 243-271. doi: 10.3934/dcdss.2020213
[6]

Christian Bläsche, Shawn Means, Carlo R. Laing. Degree assortativity in networks of spiking neurons. Journal of Computational Dynamics, 2020, 7 (2) : 401-423. doi: 10.3934/jcd.2020016
[7]

Lars Grüne. Computing Lyapunov functions using deep neural networks. Journal of Computational Dynamics, 2021, 8 (2) : 131-152. doi: 10.3934/jcd.2021006
[8]

Denis Mercier, Serge Nicaise. Existence results for general systems of differential equations on one-dimensional networks and prewavelets approximation. Discrete & Continuous Dynamical Systems, 1998, 4 (2) : 273-300. doi: 10.3934/dcds.1998.4.273
[9]

D. Lannes. Consistency of the KP approximation. Conference Publications, 2003, 2003 (Special) : 517-525. doi: 10.3934/proc.2003.2003.517
[10]

Cristina Stoica. An approximation theorem in classical mechanics. Journal of Geometric Mechanics, 2016, 8 (3) : 359-374. doi: 10.3934/jgm.2016011
[11]

Hongfei Yang, Xiaofeng Ding, Raymond Chan, Hui Hu, Yaxin Peng, Tieyong Zeng. A new initialization method based on normed statistical spaces in deep networks. Inverse Problems & Imaging, 2021, 15 (1) : 147-158. doi: 10.3934/ipi.2020045
[12]

Anne-Sophie de Suzzoni. Consequences of the choice of a particular basis of $L^2(S^3)$ for the cubic wave equation on the sphere and the Euclidean space. Communications on Pure & Applied Analysis, 2014, 13 (3) : 991-1015. doi: 10.3934/cpaa.2014.13.991
[13]

Jakub Cupera. Diffusion approximation of neuronal models revisited. Mathematical Biosciences & Engineering, 2014, 11 (1) : 11-25. doi: 10.3934/mbe.2014.11.11
[14]

Bernd Aulbach, Martin Rasmussen, Stefan Siegmund. Approximation of attractors of nonautonomous dynamical systems. Discrete & Continuous Dynamical Systems - B, 2005, 5 (2) : 215-238. doi: 10.3934/dcdsb.2005.5.215
[15]

Rua Murray. Approximation error for invariant density calculations. Discrete & Continuous Dynamical Systems, 1998, 4 (3) : 535-557. doi: 10.3934/dcds.1998.4.535
[16]

Bo Tan, Qinglong Zhou. Approximation properties of Lüroth expansions. Discrete & Continuous Dynamical Systems, 2021, 41 (6) : 2873-2890. doi: 10.3934/dcds.2020389
[17]

Janne M.J. Huttunen, J. P. Kaipio. Approximation errors in nonstationary inverse problems. Inverse Problems & Imaging, 2007, 1 (1) : 77-93. doi: 10.3934/ipi.2007.1.77
[18]

Nicolas Fournier. Particle approximation of some Landau equations. Kinetic & Related Models, 2009, 2 (3) : 451-464. doi: 10.3934/krm.2009.2.451
[19]

Michael Herty, Gabriella Puppo, Sebastiano Roncoroni, Giuseppe Visconti. The BGK approximation of kinetic models for traffic. Kinetic & Related Models, 2020, 13 (2) : 279-307. doi: 10.3934/krm.2020010
[20]

Igor E. Pritsker and Richard S. Varga. Weighted polynomial approximation in the complex plane. Electronic Research Announcements, 1997, 3: 38-44.

2020 Impact Factor: 1.916

Tools

Metrics

  • PDF downloads (183)
  • HTML views (88)
  • Cited by (0)

Other articles
by authors

[Back to Top]

Copyright © 2021 American Institute of Mathematical Sciences