Optimization of Annealed Importance Sampling Hyperparameters

被引:0
|
作者
Goshtasbpour, Shirin [1 ,2 ]
Perez-Cruz, Fernando [1 ,2 ]
机构
[1] Swiss Fed Inst Technol, Dept Comp Sci, Ramistr 101, CH-8092 Zurich, Switzerland
[2] Swiss Data Sci Ctr, Turnerstr 1, CH-8092 Zurich, Switzerland
关键词
Annealed importance sampling; Partition function estimation; Generative models; INFERENCE;
D O I
10.1007/978-3-031-26419-1_11
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
Annealed Importance Sampling (AIS) is a popular algorithm used to estimates the intractable marginal likelihood of deep generative models. Although AIS is guaranteed to provide unbiased estimate for any set of hyperparameters, the common implementations rely on simple heuristics such as the geometric average bridging distributions between initial and the target distribution which affect the estimation performance when the computation budget is limited. In order to reduce the number of sampling iterations, we present a parameteric AIS process with flexible intermediary distributions defined by a residual density with respect to the geometric mean path. Our method allows parameter sharing between annealing distributions, the use of fix linear schedule for discretization and amortization of hyperparameter selection in latent variable models. We assess the performance of Optimized-Path AIS for marginal likelihood estimation of deep generative models and compare it to compare it to more computationally intensive AIS.
引用
收藏
页码:174 / 190
页数:17
相关论文
共 50 条
  • [1] Annealed importance sampling
    Neal, RM
    [J]. STATISTICS AND COMPUTING, 2001, 11 (02) : 125 - 139
  • [2] Annealed importance sampling
    Radford M. Neal
    [J]. Statistics and Computing, 2001, 11 : 125 - 139
  • [3] Annealed importance sampling of peptides
    Lyman, Edward R.
    Zuckerman, Daniel M.
    [J]. BIOPHYSICAL JOURNAL, 2007, : 151A - 151A
  • [4] Annealed Adaptive Importance Sampling
    Center, Julian L., Jr.
    [J]. BAYESIAN INFERENCE AND MAXIMUM ENTROPY METHODS IN SCIENCE AND ENGINEERING, 2008, 1073 : 119 - 126
  • [5] Annealed importance sampling of peptides
    Lyman, Edward
    Zuckerman, Daniel M.
    [J]. JOURNAL OF CHEMICAL PHYSICS, 2007, 127 (06):
  • [6] Surrogate Likelihoods for Variational Annealed Importance Sampling
    Jankowiak, Martin
    Du Phan
    [J]. INTERNATIONAL CONFERENCE ON MACHINE LEARNING, VOL 162, 2022,
  • [7] Annealed Importance Sampling for Neural Mass Models
    Penny, Will
    Sengupta, Biswa
    [J]. PLOS COMPUTATIONAL BIOLOGY, 2016, 12 (03)
  • [8] Annealed importance sampling with constant cooling rate
    Giovannelli, Edoardo
    Cardini, Gianni
    Gellini, Cristina
    Pietraperzia, Giangaetano
    Chelli, Riccardo
    [J]. JOURNAL OF CHEMICAL PHYSICS, 2015, 142 (07):
  • [9] Iterative ensemble smoothers in the annealed importance sampling framework
    Stordal, Andreas S.
    Elsheikh, Ahmed H.
    [J]. ADVANCES IN WATER RESOURCES, 2015, 86 : 231 - 239
  • [10] Annealed Importance Sampling Reversible Jump MCMC Algorithms
    Karagiannis, Georgios
    Andrieu, Christophe
    [J]. JOURNAL OF COMPUTATIONAL AND GRAPHICAL STATISTICS, 2013, 22 (03) : 623 - 648