17643016. TRAINING ALGORITHMS FOR ONLINE MACHINE LEARNING simplified abstract (INTERNATIONAL BUSINESS MACHINES CORPORATION)

From WikiPatents
Jump to navigation Jump to search

TRAINING ALGORITHMS FOR ONLINE MACHINE LEARNING

Organization Name

INTERNATIONAL BUSINESS MACHINES CORPORATION

Inventor(s)

Dustin Michael Sargent of San Diego CA (US)

Russell L. Klenk of Julian CA (US)

Sun Young Park of San Diego CA (US)

TRAINING ALGORITHMS FOR ONLINE MACHINE LEARNING - A simplified explanation of the abstract

This abstract first appeared for US patent application 17643016 titled 'TRAINING ALGORITHMS FOR ONLINE MACHINE LEARNING

Simplified Explanation

The patent application describes a method for updating deployed AI models by concurrently deploying a pre-trained model and comparing its output to the currently deployed model. Here are the key points:

  • Currently deployed models are replaced with new models when the corresponding algorithms are updated.
  • The updated algorithm is pre-trained offline on the same training data used by the currently deployed model.
  • The pre-trained model is concurrently deployed with the currently deployed model within the same AI system.
  • The pre-trained model undergoes secondary training during the operation of the currently deployed model.
  • The output of the currently deployed model is compared to the output of the pre-trained model for the same input.
  • A decreasing rewards process encourages the pre-trained model to match the output of the currently deployed model.
  • This process continues until a condition is met, at which point the pre-trained model becomes the currently deployed model.
  • The previously deployed model is then no longer in use.

Potential applications of this technology:

  • Updating AI models in various industries such as healthcare, finance, and transportation.
  • Improving the accuracy and performance of AI systems in real-time applications.
  • Enabling seamless transitions between different versions of AI models without interrupting the system's operation.

Problems solved by this technology:

  • Ensures that AI models can be updated and improved without disrupting the system's functionality.
  • Provides a mechanism for continuous learning and improvement of AI models during their deployment.
  • Reduces the need for manual intervention in updating and replacing deployed AI models.

Benefits of this technology:

  • Allows for more efficient and effective updates of AI models in production environments.
  • Enables AI systems to adapt and improve over time without interrupting their operation.
  • Enhances the accuracy and reliability of AI systems by continuously training and updating the models.


Original Abstract Submitted

Currently deployed models are replaced with new models when corresponding deployed algorithms are updated. The updated algorithm is pre-trained offline on training data used by the currently deployed model. Concurrent deployment of the pre-trained model during operation of the currently deployed model within the same AI system provides secondary training of the pre-trained model. For the same input, output of the currently deployed model is compared to output of the pre-trained model and a decreasing rewards process encourages matching output to that of the currently deployed model until a condition is met. Upon meeting the condition, the pre-trained model become the currently deployed model and the previously deployed model is no longer in use.

Retrieved from "http://wikipatents.org/index.php?title=17643016._TRAINING_ALGORITHMS_FOR_ONLINE_MACHINE_LEARNING_simplified_abstract_(INTERNATIONAL_BUSINESS_MACHINES_CORPORATION)&oldid=25755"