Fast generalization error bound of deep learning without scale invariance of activation functions
From MaRDI portal
Publication:2055056
DOI10.1016/j.neunet.2020.05.033zbMath1475.68323arXiv1907.10900OpenAlexW3036319863WikidataQ96771349 ScholiaQ96771349MaRDI QIDQ2055056
Ryoma Hirose, Yoshikazu Terada
Publication date: 3 December 2021
Published in: Neural Networks (Search for Journal in Brave)
Full work available at URL: https://arxiv.org/abs/1907.10900
deep learningfast learning rateempirical risk minimizerexponential linear unitsigmoid activation function
Cites Work
- Unnamed Item
- Unnamed Item
- Unnamed Item
- Unnamed Item
- Unnamed Item
- Harmonic analysis of neural networks
- Sharper bounds for Gaussian and empirical processes
- Multilayer feedforward networks are universal approximators
- Empirical margin distributions and bounding the generalization error of combined classifiers
- Weak convergence and empirical processes. With applications to statistics
- Optimal approximation of piecewise smooth functions using deep ReLU neural networks
- Nonparametric regression using deep neural networks with ReLU activation function
- Error bounds for approximations with deep ReLU networks
- Local Rademacher complexities and oracle inequalities in risk minimization. (2004 IMS Medallion Lecture). (With discussions and rejoinder)
- Neural network with unbounded activation functions is universal approximator
- Statistical performance of support vector machines
- Local Rademacher complexities
- On the Expressive Power of Deep Architectures
- Support Vector Machines
- High-Dimensional Probability
- Improving the sample complexity using global data
- Breaking the Curse of Dimensionality with Convex Neural Networks
- On the Equivalence between Kernel Quadrature Rules and Random Feature Expansions
- Theory of Reproducing Kernels
- Approximation by superpositions of a sigmoidal function
- On consistency of kernel density estimators for randomly censored data: Rates holding uniformly over adaptive intervals
