AI Paper Atlas

Research map / Learning theory and algorithms / Kernels and regression theory

Quasi-Monte Carlo Feature Maps for Shift-Invariant Kernels

Haim Avron, Vikas Sindhwani, Jiyan Yang, Michael W. Mahoney

JMLR 2016 · Learning theory and algorithms › Kernels and regression theory · approach: Random feature approximation · 88 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 Halton QMC sequence instead of Monte Carlo random featuresrandom Fourier features Gaussian kernel · census, s=400regression error: 0.0791 -> 0.0529Improved
use Halton QMC sequence instead of Monte Carlo random featuresrandom Fourier features Gaussian kernel · cpu, s=100regression error: 0.0383 -> 0.0367Improved
use Sobol' sequence instead of Monte Carlorandom Fourier features Gaussian kernel · Gram matrix approximationrelative Gram errorNo clear effect
use scrambled QMC sequences instead of non-scrambled (ablation)QMC feature maps · Gram matrix approximationrelative Gram error ratioImproved
use QMC sequences instead of MC for classificationapproximate kernel ridge regression · classification datasets, s<2000accuracyNo clear effect
measure box discrepancy on central box b/2 instead of full box (ablation)box discrepancy as predictor · Gram matrix approximationcorrelation with Gram errorImproved
add adaptive (learned) QMC sequences over HaltonHalton sequence · cpu and census discrepancybox discrepancy: 3.41e-3 -> 1.29e-6Improved
learn only weights on Halton features (Weighted) instead of HaltonHalton sequence · census, s=400, D^bbox discrepancy: 2.61e-3 -> 8.83e-4Improved
use Global adaptive sequence optimized on full box instead of Haltonkernel ridge regression · census, s=400regression error: 0.0529 -> 0.1034Hurt
use Greedy adaptive (b/4) instead of Haltonkernel ridge regression · cpu, s=500regression error: 0.0348 -> 0.0291Improved
optimize adaptive sequence for the central box b/2 instead of the full box (ablation)Global adaptive on housing · housing Gram matrix errormean and norm errorMixed
run Global adaptive optimization for more iterations on housing (ablation)Global adaptive · housingGram error metricsHurt

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