Function approximation by deep networks
arXiv:1905.12882
Abstract
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.
arXiv commentTo appear in Communications in pure and applied mathematics
References in corpus (6)
- On the Number of Linear Regions of Deep Neural Networks
- Optimal approximation of continuous functions by very deep ReLU networks
- Depth-Width Tradeoffs in Approximating Natural Functions with Neural Networks
- Benefits of depth in neural networks
- Localized linear polynomial operators and quadrature formulas on the sphere
- Bounding and Counting Linear Regions of Deep Neural Networks