AI Paper Atlas

Research map / Learning theory and algorithms / Matrix and tensor methods

Local Hessian Spectral Filtering for Robust Intrinsic Dimension Estimation

GENKI OSADA

ICML 2026 · Learning theory and algorithms › Matrix and tensor methods · approach: High-dimensional estimation theory · 0 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 Hessian spectral filtering (LHSD m=5) instead of FLIPDFLIPD · Synthetic manifolds average (9 settings)MAE: 307.4 -> 6.6Improved
use LHSD instead of FLIPD in high co-dimensionFLIPD · F^10+80+200 in R^3072MAE: 1240.89 -> 5.19Improved
use LHSD instead of FLIPD on nonlinear d=900 manifoldFLIPD · F^900 in R^3072MAE: 782.5 -> 18.79Improved
increase Lanczos steps from m=2 to m=5 (ablation)LHSD · Synthetic manifolds averageMAE: 20.7 -> 6.6Improved
increase Lanczos steps from m=5 to 10 or 20 (ablation)LHSD · Synthetic manifoldsMAENo clear effect
increase Lanczos steps from m=2 to m=5 (ablation)LHSD · IDR benchmarks runtimeruntime (ms): 480 -> 1,200Hurt
use Hessian spectral filtering (LHSD m=2) instead of FLIPDFLIPD · IDR benchmark runtimeruntime (ms): 120,000 -> 480Improved
apply Hutchinson estimation (K=8) to FLIPD (ablation)FLIPD · IDR benchmarkMAE: error > 12.9Hurt
use LHSD instead of FLIPD under IDR embedding D=784FLIPD · IDR FunnelMAE: 1.606 -> 0.661Improved
train score model on 500k instead of 50k samples (ablation)LHSD · L^10+80+200 D=1024MAE: 4.16 -> 3.47Improved
train score model on 500k instead of 50k samples (ablation)LHSD · F^10+80+200 D=1024MAE: 7.52 -> 6.90Improved
vary filter steepness p (ablation)LHSD · L^900 in R^3072MAENo clear effect

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