US Patent Application 18312584. METHODS, APPARATUS, AND ARTICLES OF MANUFACTURE TO RE-PARAMETERIZE MULTIPLE HEAD NETWORKS OF AN ARTIFICIAL INTELLIGENCE MODEL simplified abstract

From WikiPatents
Jump to navigation Jump to search

METHODS, APPARATUS, AND ARTICLES OF MANUFACTURE TO RE-PARAMETERIZE MULTIPLE HEAD NETWORKS OF AN ARTIFICIAL INTELLIGENCE MODEL

Organization Name

Intel Corporation


Inventor(s)

Vinnam Kim of Pyeongtaek-si (KR)

Wonju Lee of Seongnam-si (KR)

Seok-Yong Byun of Seoul (KR)

METHODS, APPARATUS, AND ARTICLES OF MANUFACTURE TO RE-PARAMETERIZE MULTIPLE HEAD NETWORKS OF AN ARTIFICIAL INTELLIGENCE MODEL - A simplified explanation of the abstract

This abstract first appeared for US patent application 18312584 titled 'METHODS, APPARATUS, AND ARTICLES OF MANUFACTURE TO RE-PARAMETERIZE MULTIPLE HEAD NETWORKS OF AN ARTIFICIAL INTELLIGENCE MODEL

Simplified Explanation

The patent application describes a method and apparatus for re-parameterizing multiple head networks of an artificial intelligence (AI) model.

  • The invention involves training an AI model using both labeled data and pseudo-labeled data.
  • The AI model includes multiple head networks.
  • After the AI model has been trained, the invention re-parameterizes only the multiple head networks into a fully connected layer.
  • Other portions of the AI model are not re-parameterized.
  • This re-parameterization process helps optimize the performance of the AI model.
  • The invention provides a more efficient and effective way to re-parameterize multiple head networks in an AI model.


Original Abstract Submitted

Systems, apparatus, articles of manufacture, and methods are disclosed re-parameterize multiple head networks of an artificial intelligence model. An example apparatus is to train an AI model using labeled data and pseudo-labeled data, the AI model including multiple head networks. Additionally, the example apparatus is to, after the AI model has been trained, re-parameterize the multiple head networks of the AI model into a fully connected layer without re-parameterizing other portions of the AI model.