17950521. MODEL TRAINING METHOD AND APPARATUS simplified abstract (Samsung Electronics Co., Ltd.)

From WikiPatents
Jump to navigation Jump to search

MODEL TRAINING METHOD AND APPARATUS

Organization Name

Samsung Electronics Co., Ltd.

Inventor(s)

Hwidong Na of Seongnam-si (KR)

Hyohyeong Kang of Hwaseong-si (KR)

Hogyeong Kim of Daejeon (KR)

Hoshik Lee of Seongnam-si (KR)

MODEL TRAINING METHOD AND APPARATUS - A simplified explanation of the abstract

This abstract first appeared for US patent application 17950521 titled 'MODEL TRAINING METHOD AND APPARATUS

Simplified Explanation

The patent application describes a method and apparatus for training a model. The method involves acquiring recognition results from both a teacher model and a student model for a given input sequence. The student model is then trained in such a way that its recognition results are indistinguishable from those of the teacher model.

  • The method involves training a student model to mimic the recognition results of a teacher model.
  • The recognition results of the teacher model and the student model should be identical or very similar.
  • The training process aims to make the student model's recognition results indistinguishable from those of the teacher model.
  • The method can be applied to various types of models, such as machine learning or artificial intelligence models.

Potential Applications

  • Speech recognition systems: The method can be used to train student models to accurately recognize and transcribe speech, similar to a teacher model.
  • Image recognition systems: Student models can be trained to identify and classify objects in images, matching the recognition results of a teacher model.
  • Natural language processing: The method can be applied to train student models to understand and generate human-like language, based on the recognition results of a teacher model.

Problems Solved

  • Overcoming the limitations of traditional model training methods, where the student model may not achieve the same level of accuracy as the teacher model.
  • Ensuring consistency and reliability in the recognition results of different models.
  • Reducing the need for extensive manual labeling or annotation of training data.

Benefits

  • Improved model training: The method allows for more effective training of student models by leveraging the knowledge and expertise of a teacher model.
  • Enhanced accuracy: The trained student model can achieve recognition results that are on par with or even surpass those of the teacher model.
  • Time and cost savings: The method reduces the need for extensive manual labeling or annotation of training data, making the training process more efficient.


Original Abstract Submitted

A model training method and apparatus is disclosed, where the model training method acquires a recognition result of a teacher model and a recognition result of a student model for an input sequence and trains the student model such that the recognition result of the teacher model and the recognition result of the student model are not distinguished from each other.

Retrieved from "http://wikipatents.org/index.php?title=17950521._MODEL_TRAINING_METHOD_AND_APPARATUS_simplified_abstract_(Samsung_Electronics_Co.,_Ltd.)&oldid=19773"