变势流:基于能量的生成模型的新型概率框架

Junn Yong Loo, Michelle Adeline, Arghya Pal, Vishnu Monn Baskaran, Chee-Ming Ting, Raphael C. -W. Phan
{"title":"变势流:基于能量的生成模型的新型概率框架","authors":"Junn Yong Loo, Michelle Adeline, Arghya Pal, Vishnu Monn Baskaran, Chee-Ming Ting, Raphael C. -W. Phan","doi":"arxiv-2407.15238","DOIUrl":null,"url":null,"abstract":"Energy based models (EBMs) are appealing for their generality and simplicity\nin data likelihood modeling, but have conventionally been difficult to train\ndue to the unstable and time-consuming implicit MCMC sampling during\ncontrastive divergence training. In this paper, we present a novel energy-based\ngenerative framework, Variational Potential Flow (VAPO), that entirely\ndispenses with implicit MCMC sampling and does not rely on complementary latent\nmodels or cooperative training. The VAPO framework aims to learn a potential\nenergy function whose gradient (flow) guides the prior samples, so that their\ndensity evolution closely follows an approximate data likelihood homotopy. An\nenergy loss function is then formulated to minimize the Kullback-Leibler\ndivergence between density evolution of the flow-driven prior and the data\nlikelihood homotopy. Images can be generated after training the potential\nenergy, by initializing the samples from Gaussian prior and solving the ODE\ngoverning the potential flow on a fixed time interval using generic ODE\nsolvers. Experiment results show that the proposed VAPO framework is capable of\ngenerating realistic images on various image datasets. In particular, our\nproposed framework achieves competitive FID scores for unconditional image\ngeneration on the CIFAR-10 and CelebA datasets.","PeriodicalId":501347,"journal":{"name":"arXiv - CS - Neural and Evolutionary Computing","volume":"65 1","pages":""},"PeriodicalIF":0.0000,"publicationDate":"2024-07-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":"{\"title\":\"Variational Potential Flow: A Novel Probabilistic Framework for Energy-Based Generative Modelling\",\"authors\":\"Junn Yong Loo, Michelle Adeline, Arghya Pal, Vishnu Monn Baskaran, Chee-Ming Ting, Raphael C. -W. Phan\",\"doi\":\"arxiv-2407.15238\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"Energy based models (EBMs) are appealing for their generality and simplicity\\nin data likelihood modeling, but have conventionally been difficult to train\\ndue to the unstable and time-consuming implicit MCMC sampling during\\ncontrastive divergence training. In this paper, we present a novel energy-based\\ngenerative framework, Variational Potential Flow (VAPO), that entirely\\ndispenses with implicit MCMC sampling and does not rely on complementary latent\\nmodels or cooperative training. The VAPO framework aims to learn a potential\\nenergy function whose gradient (flow) guides the prior samples, so that their\\ndensity evolution closely follows an approximate data likelihood homotopy. An\\nenergy loss function is then formulated to minimize the Kullback-Leibler\\ndivergence between density evolution of the flow-driven prior and the data\\nlikelihood homotopy. Images can be generated after training the potential\\nenergy, by initializing the samples from Gaussian prior and solving the ODE\\ngoverning the potential flow on a fixed time interval using generic ODE\\nsolvers. Experiment results show that the proposed VAPO framework is capable of\\ngenerating realistic images on various image datasets. In particular, our\\nproposed framework achieves competitive FID scores for unconditional image\\ngeneration on the CIFAR-10 and CelebA datasets.\",\"PeriodicalId\":501347,\"journal\":{\"name\":\"arXiv - CS - Neural and Evolutionary Computing\",\"volume\":\"65 1\",\"pages\":\"\"},\"PeriodicalIF\":0.0000,\"publicationDate\":\"2024-07-21\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"\",\"citationCount\":\"0\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"arXiv - CS - Neural and Evolutionary Computing\",\"FirstCategoryId\":\"1085\",\"ListUrlMain\":\"https://doi.org/arxiv-2407.15238\",\"RegionNum\":0,\"RegionCategory\":null,\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"\",\"JCRName\":\"\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"arXiv - CS - Neural and Evolutionary Computing","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/arxiv-2407.15238","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
引用次数: 0

摘要

基于能量的模型(EBM)因其在数据似然建模中的通用性和简易性而备受青睐,但由于在对比发散训练过程中隐含的 MCMC 采样不稳定且耗时,因此一直难以训练。在本文中,我们提出了一种新颖的基于能量的生成框架--变异势能流(VAPO),它完全不需要隐式 MCMC 采样,也不依赖互补潜模型或合作训练。VAPO 框架旨在学习一个势能函数,该函数的梯度(流)可引导先验样本,从而使其密度演化紧跟近似数据似然同调。然后制定一个能量损失函数,以最小化流量驱动的先验样本密度演化与数据似然同构之间的库尔贝-莱伯勒差分。通过高斯先验初始化样本,并使用通用 ODE 求解器在固定时间间隔内求解支配势流的 ODE,可以在训练势能后生成图像。实验结果表明,所提出的 VAPO 框架能够在各种图像数据集上生成逼真的图像。特别是,我们提出的框架在 CIFAR-10 和 CelebA 数据集上的无条件图像生成中取得了具有竞争力的 FID 分数。
本文章由计算机程序翻译,如有差异,请以英文原文为准。
查看原文
分享 分享
微信好友 朋友圈 QQ好友 复制链接
本刊更多论文
Variational Potential Flow: A Novel Probabilistic Framework for Energy-Based Generative Modelling
Energy based models (EBMs) are appealing for their generality and simplicity in data likelihood modeling, but have conventionally been difficult to train due to the unstable and time-consuming implicit MCMC sampling during contrastive divergence training. In this paper, we present a novel energy-based generative framework, Variational Potential Flow (VAPO), that entirely dispenses with implicit MCMC sampling and does not rely on complementary latent models or cooperative training. The VAPO framework aims to learn a potential energy function whose gradient (flow) guides the prior samples, so that their density evolution closely follows an approximate data likelihood homotopy. An energy loss function is then formulated to minimize the Kullback-Leibler divergence between density evolution of the flow-driven prior and the data likelihood homotopy. Images can be generated after training the potential energy, by initializing the samples from Gaussian prior and solving the ODE governing the potential flow on a fixed time interval using generic ODE solvers. Experiment results show that the proposed VAPO framework is capable of generating realistic images on various image datasets. In particular, our proposed framework achieves competitive FID scores for unconditional image generation on the CIFAR-10 and CelebA datasets.
求助全文
通过发布文献求助,成功后即可免费获取论文全文。 去求助
来源期刊
自引率
0.00%
发文量
0
期刊最新文献
Hardware-Friendly Implementation of Physical Reservoir Computing with CMOS-based Time-domain Analog Spiking Neurons Self-Contrastive Forward-Forward Algorithm Bio-Inspired Mamba: Temporal Locality and Bioplausible Learning in Selective State Space Models PReLU: Yet Another Single-Layer Solution to the XOR Problem Inferno: An Extensible Framework for Spiking Neural Networks
×
引用
GB/T 7714-2015
复制
MLA
复制
APA
复制
导出至
BibTeX EndNote RefMan NoteFirst NoteExpress
×
×
提示
您的信息不完整,为了账户安全,请先补充。
现在去补充
×
提示
您因"违规操作"
具体请查看互助需知
我知道了
×
提示
现在去查看 取消
×
提示
确定
0
微信
客服QQ
Book学术公众号 扫码关注我们
反馈
×
意见反馈
请填写您的意见或建议
请填写您的手机或邮箱
已复制链接
已复制链接
快去分享给好友吧!
我知道了
×
扫码分享
扫码分享
Book学术官方微信
Book学术文献互助
Book学术文献互助群
群 号:481959085
Book学术
文献互助 智能选刊 最新文献 互助须知 联系我们:info@booksci.cn
Book学术提供免费学术资源搜索服务,方便国内外学者检索中英文文献。致力于提供最便捷和优质的服务体验。
Copyright © 2023 Book学术 All rights reserved.
ghs 京公网安备 11010802042870号 京ICP备2023020795号-1