Optimal approximation of continuous functions by very deep ReLU networks

Dmitry Yarotsky
Proceedings of the 31st Conference On Learning Theory, PMLR 75:639-649, 2018.

Abstract

We consider approximations of general continuous functions on finite-dimensional cubes by general deep ReLU neural networks and study the approximation rates with respect to the modulus of continuity of the function and the total number of weights $W$ in the network. We establish the complete phase diagram of feasible approximation rates and show that it includes two distinct phases. One phase corresponds to slower approximations that can be achieved with constant-depth networks and continuous weight assignments. The other phase provides faster approximations at the cost of depths necessarily growing as a power law $L\sim W^{\alpha}, 0<\alpha\le 1,$ and with necessarily discontinuous weight assignments. In particular, we prove that constant-width fully-connected networks of depth $L\sim W$ provide the fastest possible approximation rate $\|f-\widetilde f\|_\infty = O(\omega_f(O(W^{-2/\nu})))$ that cannot be achieved with less deep networks.

Cite this Paper


BibTeX
@InProceedings{pmlr-v75-yarotsky18a, title = {Optimal approximation of continuous functions by very deep ReLU networks}, author = {Yarotsky, Dmitry}, booktitle = {Proceedings of the 31st Conference On Learning Theory}, pages = {639--649}, year = {2018}, editor = {Bubeck, S├ębastien and Perchet, Vianney and Rigollet, Philippe}, volume = {75}, series = {Proceedings of Machine Learning Research}, month = {06--09 Jul}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v75/yarotsky18a/yarotsky18a.pdf}, url = {https://proceedings.mlr.press/v75/yarotsky18a.html}, abstract = {We consider approximations of general continuous functions on finite-dimensional cubes by general deep ReLU neural networks and study the approximation rates with respect to the modulus of continuity of the function and the total number of weights $W$ in the network. We establish the complete phase diagram of feasible approximation rates and show that it includes two distinct phases. One phase corresponds to slower approximations that can be achieved with constant-depth networks and continuous weight assignments. The other phase provides faster approximations at the cost of depths necessarily growing as a power law $L\sim W^{\alpha}, 0<\alpha\le 1,$ and with necessarily discontinuous weight assignments. In particular, we prove that constant-width fully-connected networks of depth $L\sim W$ provide the fastest possible approximation rate $\|f-\widetilde f\|_\infty = O(\omega_f(O(W^{-2/\nu})))$ that cannot be achieved with less deep networks. } }
Endnote
%0 Conference Paper %T Optimal approximation of continuous functions by very deep ReLU networks %A Dmitry Yarotsky %B Proceedings of the 31st Conference On Learning Theory %C Proceedings of Machine Learning Research %D 2018 %E S├ębastien Bubeck %E Vianney Perchet %E Philippe Rigollet %F pmlr-v75-yarotsky18a %I PMLR %P 639--649 %U https://proceedings.mlr.press/v75/yarotsky18a.html %V 75 %X We consider approximations of general continuous functions on finite-dimensional cubes by general deep ReLU neural networks and study the approximation rates with respect to the modulus of continuity of the function and the total number of weights $W$ in the network. We establish the complete phase diagram of feasible approximation rates and show that it includes two distinct phases. One phase corresponds to slower approximations that can be achieved with constant-depth networks and continuous weight assignments. The other phase provides faster approximations at the cost of depths necessarily growing as a power law $L\sim W^{\alpha}, 0<\alpha\le 1,$ and with necessarily discontinuous weight assignments. In particular, we prove that constant-width fully-connected networks of depth $L\sim W$ provide the fastest possible approximation rate $\|f-\widetilde f\|_\infty = O(\omega_f(O(W^{-2/\nu})))$ that cannot be achieved with less deep networks.
APA
Yarotsky, D.. (2018). Optimal approximation of continuous functions by very deep ReLU networks. Proceedings of the 31st Conference On Learning Theory, in Proceedings of Machine Learning Research 75:639-649 Available from https://proceedings.mlr.press/v75/yarotsky18a.html.

Related Material