Correcting Predictions for Approximate Bayesian Inference

Tomasz Kusmierczyk, Joseph Sakaya, Arto Klami

Tutkimustuotos: Artikkeli kirjassa/raportissa/konferenssijulkaisussaKonferenssiartikkeliTieteellinenvertaisarvioitu


Bayesian models quantify uncertainty and facilitate optimal decision-making in downstream applications. For most models, however, practitioners are forced to use approximate inference techniques that lead to sub-optimal decisions due to incorrect posterior predictive distributions. We present a novel approach that corrects for inaccuracies in posterior inference by altering the decision-making process. We train a separate model to make optimal decisions under the approximate posterior, combining interpretable Bayesian modeling with optimization of direct predictive accuracy in a principled fashion. The solution is generally applicable as a plug-in module for predictive decision-making for arbitrary probabilistic programs, irrespective of the posterior inference strategy. We demonstrate the approach empirically in several problems, confirming its potential.
OtsikkoThe thirty-fourth AAAI conference on artificial intelligence
KustantajaAssociation for the Advancement of Artificial Intelligence (AAAI)
Julkaisupäivähelmik. 2020
ISBN (elektroninen)978-1-57735-835-0
DOI - pysyväislinkit
TilaJulkaistu - helmik. 2020
OKM-julkaisutyyppiA4 Artikkeli konferenssijulkaisuussa
Tapahtuma34th AAAI Conference on Artificial Intelligence (AAAI-20) - New York, Yhdysvallat (USA)
Kesto: 7 helmik. 202012 helmik. 2020


NimiProceedings of the AAAI Conference on Artificial Intelligence
KustantajaAAAI Press
ISSN (painettu)2159-5399
ISSN (elektroninen)2374-3468


  • 113 Tietojenkäsittely- ja informaatiotieteet

Siteeraa tätä