Skip to content
PatentBrief
US 4914603Freedom to Build
Public domain since 2008

You can freely build on How to Speed Up Neural Network Training Using Momentum

This patent expired in 2008. Every claim — 0 independent, 0 dependent — is now unenforceable. Anyone can use, reproduce, manufacture, sell, or offer for sale this technology without a license.

Original assignee

GTE Laboratories Inc

Patent granted

1990

Expired

2008

Forward citations

29

What this patent covers

The patent describes a way to train neural networks by splitting each neural unit into two parts. The first part handles the primary learning, while the second part tracks changes over time to create an 'activating variable.' This variable acts like momentum; it is added to the feedback signal to accelerate how quickly the network adjusts its internal variables (weights) to reach a correct solution. By comparing the network's output to a desired target and iterating through examples, the system converges on a result faster than standard methods of the time.

What is now free to use

All 0 claims of US 4914603 are in the public domain. Specifically:

    The 0 dependent claims add narrowing limitations and are also free.

    What is NOT covered

    Patent expiry frees this specific invention. Separately-patented improvements made after expiry may still be protected.

    • Does not cover modern deep learning architectures like Transformers or Convolutional Neural Networks.

    • Does not cover hardware-specific implementations like GPUs or TPUs.

    • Does not cover unsupervised learning techniques where no desired output is provided.

    • Does not cover backpropagation algorithms that do not utilize this specific two-subunit momentum mechanism.

    Who is building on this today

    Modern AI research labs like OpenAI, Google DeepMind, and Meta AI build on the core concept of momentum-based optimization, though they use far more sophisticated mathematical variants than those described in this 1988 filing.

    Products built on expired version of this technology

    Early artificial neural network research software

    Gradient descent optimization algorithms with momentum

    How to cite this patent in your documentation

    GTE Laboratories Inc. US Patent 4914603. Training neural networks. Granted 1990, expired 2008. Now in the public domain.

    Note: This is a convenience citation. Consult a patent attorney for formal freedom-to-operate analysis.

    PatentBrief is an educational resource and does not provide legal advice. Patent expiration information is derived from USPTO records and may not reflect continuation patents, divisional filings, or separately-patented improvements. For commercial use or production decisions, obtain a formal freedom-to-operate (FTO) opinion from a registered patent attorney.

    ← Full patent detailsMore public domain patents →