Jove
Visualize
Contact Us
JoVE
x logofacebook logolinkedin logoyoutube logo
ABOUT JoVE
OverviewLeadershipBlogJoVE Help Center
AUTHORS
Publishing ProcessEditorial BoardScope & PoliciesPeer ReviewFAQSubmit
LIBRARIANS
TestimonialsSubscriptionsAccessResourcesLibrary Advisory BoardFAQ
RESEARCH
JoVE JournalMethods CollectionsJoVE Encyclopedia of ExperimentsArchive
EDUCATION
JoVE CoreJoVE BusinessJoVE Science EducationJoVE Lab ManualFaculty Resource CenterFaculty Site
Terms & Conditions of Use
Privacy Policy
Policies

Related Experiment Video

Updated: Dec 12, 2025

Super-resolution Imaging of Proteus mirabilis Biofilm by Expansion Microscopy
07:10

Super-resolution Imaging of Proteus mirabilis Biofilm by Expansion Microscopy

Published on: July 18, 2025

890

DANTE: Deep alternations for training neural networks.

Vaibhav B Sinha1, Sneha Kudugunta1, Adepu Ravi Sankar1

  • 1Department of Computer Science and Engineering, Indian Institute of Technology Hyderabad, India.

Neural Networks : the Official Journal of the International Neural Network Society
|August 11, 2020
PubMed
Summary

We introduce DANTE, a new neural network training method using alternating minimization. This approach offers a competitive alternative to backpropagation, achieving comparable results in solution quality and training speed.

Related Concept Videos

You might also read

Related Articles

Articles linked to this work by shared authors, journal, and citation graph.

Sort by
Same author

Deployment of a Machine Learning Algorithm in a Real-World Cohort for Quality Control Monitoring of Human Epidermal Growth Factor-2-Stained Clinical Specimens in Breast Cancer.

Archives of pathology & laboratory medicine·2024
Same author

How Useful Is Image-Based Active Learning for Plant Organ Segmentation?

Plant phenomics (Washington, D.C.)·2022
Same author

Incremental Object Detection via Meta-Learning.

IEEE transactions on pattern analysis and machine intelligence·2021
Same author

Active learning with point supervision for cost-effective panicle detection in cereal crops.

Plant methods·2020
Same author

Automatic estimation of heading date of paddy rice using deep learning.

Plant methods·2019

Area of Science:

  • Artificial Intelligence
  • Machine Learning
  • Optimization

Background:

  • Traditional neural network training relies heavily on gradient-based backpropagation.
  • Backpropagation faces challenges with non-differentiable activation functions and can be computationally intensive.
  • Exploring alternative training methodologies is crucial for advancing deep learning capabilities.

Purpose of the Study:

  • To introduce DANTE, a novel method for training neural networks.
  • To present an alternative to gradient-based backpropagation using alternating minimization.
  • To demonstrate the effectiveness of DANTE across various neural network architectures.

Main Methods:

  • DANTE employs the alternating minimization principle for neural network training.
Keywords:
BackpropagationDeep learningMachine learningNeural nets

Related Experiment Videos

Last Updated: Dec 12, 2025

Super-resolution Imaging of Proteus mirabilis Biofilm by Expansion Microscopy
07:10

Super-resolution Imaging of Proteus mirabilis Biofilm by Expansion Microscopy

Published on: July 18, 2025

890
  • The method frames neural network training as a bi-quasi-convex optimization problem.
  • It effectively handles both differentiable (e.g., sigmoid) and non-differentiable (e.g., ReLU) activation functions.

    • Main Results:

    • Neural networks trained with DANTE show promising and competitive performance compared to backpropagation.
    • The method achieves comparable solution quality to traditional techniques.
    • DANTE demonstrates competitive training speeds on standard datasets.

    Conclusions:

    • DANTE offers a viable and effective alternative to backpropagation for training neural networks.
    • The method's ability to handle diverse activation functions and its competitive performance highlight its potential.
    • Further research into DANTE could lead to advancements in deep learning training efficiency and effectiveness.