On the Expressive Power of Neural Networks
This work addresses foundational theoretical problems in machine learning regarding the capabilities of neural network architectures, but it appears incremental as it builds on prior approximation theorems.
The paper tackles unresolved questions about the expressive power of neural networks, such as whether wide shallow ReLU-networks can be approximated by deep narrow ones and if universal approximation holds for other norms and activation functions, by introducing a new framework with two expressive powers and improving bounds on linear regions.
In 1989 George Cybenko proved in a landmark paper that wide shallow neural networks can approximate arbitrary continuous functions on a compact set. This universal approximation theorem sparked a lot of follow-up research. Shen, Yang and Zhang determined optimal approximation rates for ReLU-networks in $L^p$-norms with $p \in [1,\infty)$. Kidger and Lyons proved a universal approximation theorem for deep narrow ReLU-networks. Telgarsky gave an example of a deep narrow ReLU-network that cannot be approximated by a wide shallow ReLU-network unless it has exponentially many neurons. However, there are even more questions that still remain unresolved. Are there any wide shallow ReLU-networks that cannot be approximated well by deep narrow ReLU-networks? Is the universal approximation theorem still true for other norms like the Sobolev norm $W^{1,1}$? Do these results hold for activation functions other than ReLU? We will answer all of those questions and more with a framework of two expressive powers. The first one is well-known and counts the maximal number of linear regions of a function calculated by a ReLU-network. We will improve the best known bounds for this expressive power. The second one is entirely new.