AI Paper Atlas

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

Autoregressive Score Matching

Chenlin Meng, Lantao Yu, Yang Song, Jiaming Song, Stefano Ermon

NeurIPS 2020 · Probabilistic and scientific modeling › Variational and generative sampling · approach: Posterior sampling with diffusion guidance · 2 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
train implicit VAE encoder with CSM instead of ELBOVAE · MNIST latent dim 8negative log-likelihood (AIS): 96.74 -> 95.02Improved
train implicit VAE encoder with CSM instead of ELBOVAE · MNIST latent dim 16negative log-likelihood (AIS): 91.82 -> 88.42Improved
train implicit VAE encoder with CSM instead of ELBOVAE · CelebA latent dim 32FID: 66.31 -> 62.20Improved
use CSM instead of sliced score matching (SSM) for the entropy gradientVAE implicit encoder · MNIST latent dim 8negative log-likelihood (AIS): 95.61 -> 95.02Improved
use CSM instead of SSMVAE implicit encoder · CelebAFID: 62.50 -> 62.20Improved
use CSM instead of Stein or Spectral estimatorsVAE implicit encoder · CelebAFID: 108.84 -> 62.20Improved
train the autoregressive score model with CSM instead of SSM (SSM-AR)AR score model for entropy gradient · CelebAFID: 66.88 -> 62.20Improved
use CSM instead of original score matching on Gaussiansscore matching · multivariate Gaussians, increasing dimensiontime per iteration and memoryImproved
use CSM instead of sliced score matchingscore matching · 100-d Gaussian N(0, 0.1^2 I)training convergenceImproved
use CSM instead of denoising score matching with small sigmascore matching · 100-d Gaussian and MNIST MADEconvergence and NLLImproved
decrease DSM noise level sigmadenoising score matching · MNIST MADEbias vs convergence speedMixed
train MADE with CSM (AR-CSM) instead of MLEMADE with Gaussian densities · MNIST/CIFAR-10/CelebA samplessample qualityImproved

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