Title:
NEURAL NETWORK COMPUTATION METHOD AND NEURAL NETWORK WEIGHT GENERATION METHOD
Document Type and Number:
WIPO Patent Application WO/2022/097954
Kind Code:
A1
Abstract:
Disclosed is a neural network driving method comprising a shifting step of obtaining one set of first values by multiplying one set of weights by corresponding values of one set of input activations, respectively, obtaining a second value by adding the one set of first values to one another, and obtaining an output activation of a first node by right-shifting the second value by n bits. Each of the one set of input activations and the one set of weights corresponds to n-bit integer-type data.
More Like This:
Inventors:
JUNG TAE-YOUNG (KR)
Application Number:
PCT/KR2021/014367
Publication Date:
May 12, 2022
Filing Date:
October 15, 2021
Export Citation:
Assignee:
OPENEDGES TECH INC (KR)
International Classes:
G06N3/063; G06N3/04; G06N3/08
Foreign References:
| KR20190062129A | 2019-06-05 | |||
| KR20190074938A | 2019-06-28 | |||
| JP2019185134A | 2019-10-24 | |||
| KR20200061164A | 2020-06-02 | |||
| KR20030072860A | 2003-09-19 |
Attorney, Agent or Firm:
YANG, Ki-Hyuk (KR)
Download PDF: