18479723. TEST-TIME ADAPTATION VIA SELF-DISTILLED REGULARIZATION simplified abstract (QUALCOMM Incorporated)

From WikiPatents
Jump to navigation Jump to search

TEST-TIME ADAPTATION VIA SELF-DISTILLED REGULARIZATION

Organization Name

QUALCOMM Incorporated

Inventor(s)

Junha Song of Daejeon (KR)

Sungha Choi of Goyang-si (KR)

TEST-TIME ADAPTATION VIA SELF-DISTILLED REGULARIZATION - A simplified explanation of the abstract

This abstract first appeared for US patent application 18479723 titled 'TEST-TIME ADAPTATION VIA SELF-DISTILLED REGULARIZATION

Simplified Explanation

The abstract describes a computer-implemented method that involves adding auxiliary networks to partitions associated with a main network, training these auxiliary networks with training data, adapting them with test data, and classifying inputs based on the adapted networks.

  • The method involves adding auxiliary networks to partitions of a main network.
  • The auxiliary networks are trained with training data to adapt to a test distribution.
  • The auxiliary networks are further adapted with test data to better fit the test distribution.
  • Inputs received at a model are classified based on the adapted auxiliary networks.

Potential Applications

This technology could be applied in various fields such as image recognition, natural language processing, and financial forecasting.

Problems Solved

This technology helps improve the adaptability and accuracy of neural networks to different distributions of data, enhancing their performance in various tasks.

Benefits

The method allows for better adaptation of neural networks to new data distributions, leading to improved classification accuracy and performance.

Potential Commercial Applications

This technology could be valuable in industries such as healthcare, finance, and e-commerce for tasks like medical image analysis, fraud detection, and personalized recommendations.

Possible Prior Art

One possible prior art could be techniques for domain adaptation in machine learning, where models are adapted to new domains with limited labeled data.

Unanswered Questions

How does this method compare to existing techniques for domain adaptation in neural networks?

This article does not provide a direct comparison to existing techniques for domain adaptation in neural networks.

What are the computational requirements of implementing this method on large-scale datasets?

This article does not address the computational requirements of implementing this method on large-scale datasets.


Original Abstract Submitted

A computer-implemented method includes adding an auxiliary network of a group of auxiliary networks to a respective partition of a group of partitions associated with a main network. The method also includes training each of the group of auxiliary networks with training data to adapt to a test distribution. The method further includes adapting each of the group of auxiliary networks with test data to adapt to the test distribution. The method still further includes classifying an input received at a model based on adapting each of the group of auxiliary networks. The model may include the group of partitions and the group of auxiliary networks.