Mining gold from implicit models to improve likelihood-free inference.
implicit models
neural density estimation
simulation-based inference
Journal
Proceedings of the National Academy of Sciences of the United States of America
ISSN: 1091-6490
Titre abrégé: Proc Natl Acad Sci U S A
Pays: United States
ID NLM: 7505876
Informations de publication
Date de publication:
10 03 2020
10 03 2020
Historique:
pubmed:
23
2
2020
medline:
23
2
2020
entrez:
22
2
2020
Statut:
ppublish
Résumé
Simulators often provide the best description of real-world phenomena. However, the probability density that they implicitly define is often intractable, leading to challenging inverse problems for inference. Recently, a number of techniques have been introduced in which a surrogate for the intractable density is learned, including normalizing flows and density ratio estimators. We show that additional information that characterizes the latent process can often be extracted from simulators and used to augment the training data for these surrogate models. We introduce several loss functions that leverage these augmented data and demonstrate that these techniques can improve sample efficiency and quality of inference.
Identifiants
pubmed: 32079725
pii: 1915980117
doi: 10.1073/pnas.1915980117
pmc: PMC7071889
doi:
Types de publication
Journal Article
Research Support, U.S. Gov't, Non-P.H.S.
Research Support, Non-U.S. Gov't
Langues
eng
Sous-ensembles de citation
IM
Pagination
5242-5249Déclaration de conflit d'intérêts
The authors declare no competing interest.
Références
Stat Comput. 2018;28(2):411-425
pubmed: 31997856
Proc Natl Acad Sci U S A. 2003 Dec 23;100(26):15324-8
pubmed: 14663152
Phys Rev Lett. 2018 Sep 14;121(11):111801
pubmed: 30265125
Genetics. 2002 Dec;162(4):2025-35
pubmed: 12524368
Proc Natl Acad Sci U S A. 2020 Mar 10;117(10):5242-5249
pubmed: 32079725
Proc Natl Acad Sci U S A. 1920 Jul;6(7):410-5
pubmed: 16576509
Proc Natl Acad Sci U S A. 2007 Feb 6;104(6):1760-5
pubmed: 17264216