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 Videos

Exploiting inherent relationships in RNN architectures.

D P. Mandic1, J A. Chambers

  • 1Communications and Signal Processing Group, Department of Electrical and Electronic Engineering, Imperial College of Science, Technology and Medicine, Exhibition Road, London, UK

Neural Networks : the Official Journal of the International Neural Network Society
|March 29, 2003
PubMed
Summary

This study reveals a link between learning rate and neural network activation slopes in Recurrent Neural Network (RNN) systems. This simplifies computational complexity by reducing optimizable parameters in adaptive learning algorithms.

Related Concept Videos

You might also read

Related Articles

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

Sort by
Same journal

Injecting reasoning into vision-language models via weight-decomposed merging.

Neural networks : the official journal of the International Neural Network Society·2026
Same journal

A Non-equilibrium thermodynamic framework for neural networks: A principled correspondence and parameter dynamics.

Neural networks : the official journal of the International Neural Network Society·2026
Same journal

Dual-pathway mask ranking guided selective fine-tuning for backdoor purification.

Neural networks : the official journal of the International Neural Network Society·2026
Same journal

Discriminative transfer feature learning for unsupervised domain adaptation.

Neural networks : the official journal of the International Neural Network Society·2026
Same journal

ARetinex-Net: Low-light image enhancement with adaptive retinex model and global illumination representation.

Neural networks : the official journal of the International Neural Network Society·2026
Same journal

A Multi-Task Learning Framework with Physics Embedded for Signal Reconstruction and State Prediction in Underground Multi-Robot Systems.

Neural networks : the official journal of the International Neural Network Society·2026

Area of Science:

  • Computational neuroscience
  • Machine learning theory
  • Artificial neural networks

Background:

  • Recurrent Neural Networks (RNNs) are foundational for sequential data processing.
  • Nonlinear modular cascaded systems offer enhanced modeling capabilities.
  • Understanding the interplay between learning rates and activation functions is crucial for efficient training.

Purpose of the Study:

  • To establish the relationship between learning rate and the slope of nonlinear activation functions in modular cascaded RNNs.
  • To investigate the impact on computational complexity and parameter optimization.
  • To analyze the behavior of Gradient Descent (GD) and Extended Recursive Least Squares (ERLS) algorithms.

Main Methods:

  • Analysis within the framework of nonlinear modular cascaded systems.

Related Experiment Videos

  • Utilizing Recurrent Neural Network (RNN) architectures.
  • Application of Gradient Descent (GD) and Extended Recursive Least Squares (ERLS) algorithms with a general nonlinear activation function.

    • Main Results:

    • A direct relationship is identified between the learning rate and the slope of the neuron's nonlinear activation function.
    • Reduced computational complexity in adaptive learning algorithms due to fewer independent parameters.
    • Degeneration of results to single RNNs when the cascaded system has only one module.

    Conclusions:

    • The findings simplify the optimization of weights in adaptive learning algorithms for cascaded RNNs.
    • The study provides a theoretical basis for efficient training of complex neural network architectures.
    • The results are generalizable and reduce to known outcomes for simpler RNN configurations.