AI Paper Atlas

Research map / Probabilistic and scientific modeling / Uncertainty and Gaussian processes

Low-rank Variational Bayes correction to the Laplace method

Janet van Niekerk, Haavard Rue

JMLR 2024 · Probabilistic and scientific modeling › Uncertainty and Gaussian processes · approach: Simulation-based inference

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
apply low-rank VBC correction to the Laplace methodLaplace method · Low-count Poisson regression (n=20)posterior mean of beta_0 (vs MCMC -0.715): -0.579 -> -0.706Improved
apply low-rank VBC correction to the Laplace methodLaplace method · Low-count Poisson regression (n=100)posterior mean of beta_0 (vs MCMC -1.196): -1.073 -> -1.199Improved
increase data size from n=20 to n=100 for VBC correctionVBC · Low-count Poisson regressionsize of correction: correction space stays 2 while parameter space grows from 22 to 102No clear effect
VBC instead of MCMC for n=100VBC · Low-count Poisson regression (n=100)time (s): 384.12 -> 17.36Improved
VBC instead of HMC for n=100VBC · Low-count Poisson regression (n=100)time (s): 169.57 -> 17.36Improved
add VBC correction on top of Laplace method for small nVBC · Low-count Poisson regression (n=20)time (s): 2.21 -> 5.78Hurt
add INLA-VBC correction to the Gaussian strategyGaussian strategy · Overdispersed Poisson regression (n=1000)posterior mean of beta_1 (vs MCMC -0.529): -0.484 -> -0.531Improved
INLA-VBC instead of INLA Laplace strategyINLA-VBC · Overdispersed Poisson regression (n=1000)time (s): 18.299 -> 5.718Improved
INLA-VBC instead of MCMCINLA-VBC · Overdispersed Poisson regression (n=1000)time (s): 207.445 -> 5.718Improved
add INLA-VBC correction to the Gaussian strategyGaussian strategy · Tokyo rainfall (cyclic RW2)mean absolute error of posterior mean vs INLA: 0.0358 -> 0.0009Improved
INLA-VBC instead of INLA on a small modelINLA-VBC · Tokyo rainfall (cyclic RW2)computational cost: all methods under 6 seconds; MCMC 87.63 secondsNo clear effect
INLA-VBC instead of INLA on a large modelINLA-VBC · Leukemia survival (Cox, spatial SPDE)time (s): 1276 -> 26.3Improved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy