Login| Sign Up| Help| Contact|

Patent Searching and Data


Title:
METHOD AND APPARATUS FOR NEURAL NETWORK BASED ON ENERGY-BASED LATENT VARIABLE MODELS
Document Type and Number:
WIPO Patent Application WO/2022/077345
Kind Code:
A1
Abstract:
Methods and apparatuses for training neural networks based on energy-based latent variable models (EBLVMs) are provided. The method comprises bi-level optimizations based on a score matching objective. The lower-level optimizes a variational posterior distribution of the latent variables to approximate the true posterior distribution of the EBLVM, and the higher-level optimizes the neural network parameters based on a modified SM objective as a function of the variational posterior distribution. The method may be applied to train neural networks based on EBLVMs with nonstructural assumptions.

Inventors:
ZHU JUN (CN)
BAO FAN (CN)
LI CHONGXUAN (CN)
XU KUN (CN)
SU HANG (CN)
LU SILIANG (CN)
Application Number:
PCT/CN2020/121172
Publication Date:
April 21, 2022
Filing Date:
October 15, 2020
Export Citation:
Click for automatic bibliography generation   Help
Assignee:
BOSCH GMBH ROBERT (DE)
UNIV TSINGHUA (CN)
International Classes:
G06N3/08
Foreign References:
US20060034495A12006-02-16
US20160300134A12016-10-13
Other References:
XIE JIANWEN, LU YANG, GAO RUIQI, WU YING NIAN: "Cooperative Learning of Energy-Based Model and Latent Variable Model via MCMC Teaching", THE THIRTY-SECOND AAAI CONFERENCE ON ARTIFICIAL INTELLIGENCE, 29 April 2018 (2018-04-29), XP055921839, Retrieved from the Internet
CHONGXUAN LI; CHAO DU; KUN XU; MAX WELLING; JUN ZHU; BO ZHANG: "To Relieve Your Headache of Training an MRF, Take AdVIL", ARXIV.ORG, CORNELL UNIVERSITY LIBRARY, 201 OLIN LIBRARY CORNELL UNIVERSITY ITHACA, NY 14853, 24 January 2019 (2019-01-24), 201 Olin Library Cornell University Ithaca, NY 14853 , XP081493492
DU YILUN, MORDATCH IGOR: "Implicit Generation and Modeling with Energy-Based Models", NEURIPS 2019, 30 June 2020 (2020-06-30), XP055921843, Retrieved from the Internet
Attorney, Agent or Firm:
NTD PATENT AND TRADEMARK AGENCY LTD. (CN)
Download PDF: