Loading...

Bounds on the approximation power of feed forward neural networks

Mehrabi, M ; Sharif University of Technology | 2018

552 Viewed
  1. Type of Document: Article
  2. Publisher: International Machine Learning Society (IMLS) , 2018
  3. Abstract:
  4. The approximation power of general feedforward neural networks with piecewise linear activation functions is investigated. First, lower bounds on the size of a network are established in terms of the approximation error and network depth and width. These bounds improve upon state- of-the-art bounds for certain classes of functions, such as strongly convex functions. Second, an upper bound is established on the difference of two neural networks with identical weights but different activation functions. © The Author(s) 2018
  5. Keywords:
  6. Chemical activation ; Functions ; Learning systems ; Piecewise linear techniques ; Activation functions ; Approximation errors ; Convex functions ; Lower bounds ; Network depths ; Piecewise linear activations ; State of the art ; Upper Bound ; Feedforward neural networks
  7. Source: 35th International Conference on Machine Learning, ICML 2018, 10 July 2018 through 15 July 2018 ; Volume 8 , 2018 , Pages 5531-5539 ; 9781510867963 (ISBN)
  8. URL: https://arxiv.org/abs/1806.11416