Title:
ADVERSARIAL LEARNING APPARATUS AND METHOD FOR SIMULTANEOUSLY TRAINING DENOISING NETWORK AND DEEP NEURAL NETWORK, AND COMPUTER-READABLE RECORDING MEDIUM HAVING RECORDED THEREON PROGRAM FOR EXECUTING METHOD
Document Type and Number:
WIPO Patent Application WO/2024/025152
Kind Code:
A1
Abstract:
The present invention relates to a technology related to an adversarial learning apparatus, and to an adversarial learning apparatus and method for simultaneously training a denoising network and a deep neural network, and a computer-readable recording medium having recorded thereon a program for executing the method, wherein the deep neural network that classifies images can maintain classification accuracy for normal images, which are reconstructed through the denoising network, at a similar level to the related art, and can also correctly classify adversarial examples reconstructed through the denoising network.
Inventors:
CHOI DAESEON (KR)
RYU GWONSANG (KR)
RYU GWONSANG (KR)
Application Number:
PCT/KR2023/008459
Publication Date:
February 01, 2024
Filing Date:
June 19, 2023
Export Citation:
Assignee:
FOUNDATION SOONGSIL UNIV INDUSTRY COOPERATION (KR)
International Classes:
G06N3/08; G06N3/045; G06N20/00; G06T5/00; G06V10/764; G06V10/82
Foreign References:
| CN110717522A | 2020-01-21 | |||
| KR102395244B1 | 2022-05-09 | |||
| KR20220030635A | 2022-03-11 | |||
| CN113988293A | 2022-01-28 |
Other References:
RYU, GWONSANG ET AL.: "HAT: Hybrid Adversarial Training for Simultaneous Training of Deep Learning Model and Denoising Network", WISA 2021, pages 20210811 - 20210813
Attorney, Agent or Firm:
YUN, Kuisang (KR)
Download PDF: