Login| Sign Up| Help| Contact|

Patent Searching and Data


Title:
TENSOR-BASED CONTINUAL LEARNING METHOD AND DEVICE
Document Type and Number:
WIPO Patent Application WO/2023/231887
Kind Code:
A1
Abstract:
The present application discloses a tensor-based continual learning method and device. The method comprises: acquiring input data, the input data comprising one or more of a video, an image, a text or a voice; and inputting the input data into a first neural network to obtain a data processing result, wherein the first neural network is obtained by training m tasks, after the training of an i-th task is ended, the neural network comprises A tensor cores, the A tensor cores are divided into B tensor layers, and each of the B tensor layers comprises data of each of the A tensor cores in a same dimension; in the training of an (i+1)-th task, C tensor cores and/or D tensor layers are added to the first neural network, and parameters in the C tensor cores and/or the D tensor layers are updated. According to the present application, the anti-forgetting capability of the model can be effectively improved, and the scale increase of the model is relatively small, so that the storage and communication overhead is effectively saved.

Inventors:
LI YINCHUAN (CN)
SHAO YUNFENG (CN)
Application Number:
PCT/CN2023/096249
Publication Date:
December 07, 2023
Filing Date:
May 25, 2023
Export Citation:
Click for automatic bibliography generation   Help
Assignee:
HUAWEI TECH CO LTD (CN)
International Classes:
G06N3/08; G06N3/04; G06N20/00
Foreign References:
CN115169548A2022-10-11
CN114463605A2022-05-10
CN113792874A2021-12-14
CN113095486A2021-07-09
US20190080222A12019-03-14
Attorney, Agent or Firm:
SCIHEAD IP LAW FIRM (CN)
Download PDF: