Global convergence and geometric characterization of slow to fast weight evolution in neural network training for classifying linearly non-separable data

DOI10.3934/ipi.2020077zbMath1469.90113arXiv2002.12563OpenAlexW3116838285MaRDI QIDQ2028920

Penghang Yin, Jack X. Xin, Ziang Long

Publication date: 3 June 2021

Published in: Inverse Problems and Imaging (Search for Journal in Brave)

Full work available at URL: https://arxiv.org/abs/2002.12563

zbMATH Keywords

classificationneural networksgradient descentlearning dynamicsgeometric conditionslow-to-fast convergence

Mathematics Subject Classification ID

Analysis of algorithms (68W40) Nonconvex programming, global optimization (90C26)

Related Items (1)

DeepParticle: learning invariant measure by a deep neural network minimizing Wasserstein distance on data generated from an interacting particle method

Uses Software

• ImageNet
• AlexNet
• CosFace
• LeNet
• Faster R-CNN

Cites Work

• Linear feature transform and enhancement of classification on deep neural network
• Unnamed Item

This page was built for publication: Global convergence and geometric characterization of slow to fast weight evolution in neural network training for classifying linearly non-separable data