An optimized method for variational autoencoders based on Gaussian cloud model

作者全名："Dai, Jin; Guo, Qiuyan; Wang, Guoyin; Liu, Xiao; Zheng, Zhifang"

作者地址："[Dai, Jin; Guo, Qiuyan; Wang, Guoyin; Liu, Xiao; Zheng, Zhifang] Chongqing Univ Posts & Telecommun, Chongqing Key Lab Computat Intelligence, Chongqing, Peoples R China; [Dai, Jin] Chongqing Univ Posts & Telecommun, Sch Software Engn, Chongqing, Peoples R China; [Guo, Qiuyan] Chongqing Med Univ, Affiliated Hosp 1, Chongqing, Peoples R China"

通信作者："Guo, QY (通讯作者)，Chongqing Univ Posts & Telecommun, Chongqing Key Lab Computat Intelligence, Chongqing, Peoples R China.; Guo, QY (通讯作者)，Chongqing Med Univ, Affiliated Hosp 1, Chongqing, Peoples R China."

来源：INFORMATION SCIENCES

ESI学科分类：COMPUTER SCIENCE

WOS号：WOS:001040047100001

JCR分区：Q1

影响因子：8.1

年份：2023

卷号：645

期号：　

开始页：　

结束页：　

文献类型：Article

关键词：Variational autoencoders; Gaussian cloud model; Loss function; Image generation

摘要："Variational Autoencoders is one of the most valuable generative models in the field of unsupervised learning. Due to its own construction characteristics, Variational Autoencoders has insufficient precision for high-resolution image reconstruction. In this paper, the priori variant model of Variational Autoencoders based on the Gaussian Cloud Model is proposed to optimize the sampling method of latent variables, network structure and loss function. First, the Gaussian Cloud Model is used to replace the prior distribution of Variational Autoencoders. Second, the sampling process is changed into two consecutive Gaussian distributions. Finally, a new loss function based on the envelope curve of the Gaussian Cloud Model is presented for approximating the real data distribution. The method is evaluated qualitatively and quantitatively on several datasets to fully demonstrate the correctness and effectiveness of the method."

基金机构：National Natural Science Foundation of China [61936001]; Natural Science Foundation Project of CQ CSTC [cstc2021jcyj-msxmX0849]

基金资助正文："& nbsp;This work has received funding in part from the National Natural Science Foundation of China under Grant No.61936001, the Natural Science Foundation Project of CQ CSTC under Grant No.cstc2021jcyj-msxmX0849."