AI Paper Atlas

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

Intrinsic dimensionality estimation using Normalizing Flows

Christian Horvat, Jean-Pascal Pfister

NeurIPS 2022 · Probabilistic and scientific modeling › Variational and generative sampling · approach: Normalizing flow training and design · 1 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 ID-NF instead of LIDL on correlated-on-sphere dataID-NF · Synthetic sphereestimated ID (true 2): 1.66±0.07 -> 2.04Improved
use ID-NF instead of LIDL on correlated swiss rollID-NF · Synthetic swiss rollestimated ID (true 2): 2.48±0.03 -> 2.02Improved
use ID-NF instead of LIDL on mixture swiss rollID-NF · Synthetic swiss rollestimated ID (true 2): 2.26±0.03 -> 2.02Improved
use ID-NF instead of twoNN on low-dimensional synthetic dataID-NF · Synthetic manifolds D<=4estimated ID: 1.98 vs 2.01 (mixture on sphere)No clear effect
use ID-NF instead of twoNN in high embedding dimension (D up to 400)ID-NF · Sphere S(D/2) in R^D, 10^4 samplesestimated IDImproved
reduce training set from 10^4 to 10^3 samples, ID-NF vs LIDLID-NF · Sphere S(D/2) in R^D, 10^3 samplesestimated IDImproved
use a single shared sigma^2 range for LIDL across distributions (no per-distribution tuning)LIDL · Synthetic manifoldsestimated ID: 1.66±0.07 on correlated sphereHurt
vary LIDL inflation-noise range on StyleGan imagesLIDL · StyleGan2 d=2/d=64estimated IDHurt
use sigma_max = 0.68*255 and only 3 trained flows on imagesID-NF · StyleGan2 d=2estimated ID (true 2): 4.06Mixed
use sigma_max = 0.68*255 and only 3 trained flows on imagesID-NF · StyleGan2 d=64estimated ID (true 64): 62.24Improved
train DNF with 130 latent dims from ID-NF instead of 512DNF · CelebA-HQFID: 34.14 -> 36.92Mixed
compare DNF with 130 latent dims against M-flow with 512DNF (130 dims) · CelebA-HQFID: 38.07 -> 36.92Improved

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