AI Paper Atlas

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

Bayesian Multinomial Logistic Normal Models through Marginally Latent Matrix-T Processes

Justin D. Silverman, Kimberly Roche, Zachary C. Holmes, Lawrence A. David, Sayan Mukherjee

JMLR 2022 · Probabilistic and scientific modeling › Uncertainty and Gaussian processes · approach: Structured Gaussian process models

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
replace HMC Collapsed with Laplace approximation (LA Collapsed)CU sampler for MLN models · Crohn's disease microbiome (83-sample subset)run time: 30 minutes -> 3 secondsImproved
add collapse step (sample marginal LTP form, then uncollapse) to HMC (ablation)HMC on MLN linear model · Simulated MLN linear modelsrun time: speed ups of between 1 to 5 orders of magnitude (LA Collapsed vs HMC)Improved
collapse the model before HMC at large N (ablation)HMC Collapsed vs HMC Uncollapsed · Simulated MLN linear models (large N)run time: HMC Uncollapsed more efficient for large NHurt
use Laplace approximation instead of HMC for collapsed formLA Collapsed · Simulated MLN linear modelspoint estimate RMSE: nearly identical point estimates over the full spectrumNo clear effect
use Laplace approximation for uncertaintyLA Collapsed · Simulated MLN linear models (Q>=250, >90% zeros)RMSE of standard deviations: identical for Q<250; differs at Q >= 250 where zeros >90%Hurt
use Laplace approximation instead of mean-field VB on the collapsed formLA Collapsed vs VB Collapsed · Simulated MLN linear modelspoint estimate and uncertainty error: 1-2 orders of magnitude faster; VB larger errors for large DImproved
use multinomial likelihood instead of pseudo-count linear model (PCLM)MLN linear model · Simulated MLN linear modelsRMSEImproved

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