Why rectified linear activation functions? Why max-pooling? A possible explanation

DOI10.1007/978-3-031-08266-5_28MaRDI QIDQ2101281zbMATH OpenOpenAlexFDO

Authors Julio C. Urenda, Vladik Kreinovich

Publication date 5 December 2022

Full work available at URL https://doi.org/10.1007/978-3-031-08266-5_28

Mathematics Subject Classification ID

Artificial neural networks and deep learning (68T07)

Recommendations

• Optimization under uncertainty explains empirical success of deep learning heuristics
• MorphoActivation: generalizing ReLU activation function by mathematical morphology
• Nonlinear approximation and (deep) ReLU networks
• Deep vs. shallow networks: an approximation theory perspective
• Spline representation and redundancies of one-dimensional ReLU neural network models

Cites work

• A first course in fuzzy logic
• Deep learning
• Fuzzy logic and mathematics: a historical perspective
• Fuzzy sets
• scientific article; zbMATH DE number 1183717 (Why is no real title available?)
• scientific article; zbMATH DE number 1354791 (Why is no real title available?)
• Uncertain Rule-Based Fuzzy Systems

Cited in

(4)

• Optimization under uncertainty explains empirical success of deep learning heuristics
• Dying ReLU and initialization: theory and numerical examples
• MorphoActivation: generalizing ReLU activation function by mathematical morphology
• Replacing pooling functions in convolutional neural networks by linear combinations of increasing functions

This page was built for publication: Why rectified linear activation functions? Why max-pooling? A possible explanation

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q2101281)