Quantized convolutional neural networks through the lens of partial differential equations

DOI10.1007/s40687-022-00354-yOpenAlexW3198630114WikidataQ114218934 ScholiaQ114218934MaRDI QIDQ2079526

Gil Ben Shalom, Moshe Eliasof, Ido Ben-Yair, Eran Treister

Publication date: 30 September 2022

Published in: Research in the Mathematical Sciences (Search for Journal in Brave)

Full work available at URL: https://arxiv.org/abs/2109.00095

zbMATH Keywords

total variationquantizationconvolutional neural networksneural ODEsstable symmetric architectures

Mathematics Subject Classification ID

Artificial neural networks and deep learning (68T07) Pattern recognition, speech recognition (68T10) Machine vision and scene understanding (68T45)

Uses Software

• CIFAR
• DeepFool
• ImageNet
• Adam
• GitHub
• U-Net
• AlexNet
• ANODE

Cites Work

• Unnamed Item
• Unnamed Item
• Unnamed Item
• Nonlinear total variation based noise removal algorithms
• Translating numerical concepts for PDEs into neural architectures
• Forward stability of ResNet and its variants
• Deep neural networks motivated by partial differential equations
• Designing rotationally invariant neural networks from PDEs and variational methods
• Deep relaxation: partial differential equations for optimizing deep neural networks
• Blended coarse gradient descent for full quantization of deep neural networks
• A proposal on machine learning via dynamical systems
• Determining optical flow
• Approximation of functional depending on jumps by elliptic functional via t-convergence
• Active contours without edges
• Stable architectures for deep neural networks
• Optimization Methods for Large-Scale Machine Learning
• Layer-Parallel Training of Deep Residual Neural Networks