AI Paper Atlas

Research map / Probabilistic and scientific modeling / Variational and generative sampling

Markovian Score Climbing: Variational Inference with KL(p||q)

Christian Naesseth, Fredrik Lindsten, David M. Blei

NeurIPS 2020 · Probabilistic and scientific modeling › Variational and generative sampling · approach: Variational inference theory and gradients · 14 citations (OpenAlex)

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
use the prior as the proposal instead of the variational approximation (ablation)MSC · Pimatest error: 0.227\pm 0.046 -> 0.456\pm 0.093Hurt
use the prior as the proposal instead of the variational approximation (ablation)MSC · Ionostest error: 0.117 \pm 0.053 -> 0.182 \pm 0.070Hurt
use the prior as the proposal instead of the variational approximation (ablation)MSC · Hearttest error: 0.160 \pm 0.063 -> 0.342 \pm 0.11Hurt
use MSC rather than self-normalized IS gradientsBayesian probit regression · Pima, Ionos, and Hearttest errorNo clear effect
use MSC rather than self-normalized IS gradientsGaussian approximation to a skew-normal posterior · skew-normal toy modelvariational parameter accuracyImproved
use MSC rather than self-normalized IS gradientsBayesian probit regression · Heart and Ionosrobustness to sample count SImproved
use subset average likelihoods for exchangeable data (ablation)inclusive-KL VI · exchangeable datatarget posterior fidelityHurt

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