AI Paper Atlas

Research map / Optimization / Nonconvex and smooth optimization

Analysis of Krylov Subspace Solutions of Regularized Non-Convex Quadratic Problems

Yair Carmon, John C. Duchi

NeurIPS 2018 · Optimization › Nonconvex and smooth optimization · approach: Adaptive first-order convex methods

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
increase Krylov subspace dimension to about 20 Lanczos iterationsKrylov subspace solution · 5,000 random cubic-regularization problems, d=10^6relative accuracy: about 10% accuracy at 20 iterationsImproved
increase Krylov subspace dimension to 100 iterationsKrylov subspace solution · 5,000 random cubic-regularization problems, d=10^6relative accuracy: better than 1% at 100 iterationsImproved
larger subspace dimension t at or above sqrt of condition numberKrylov subspace solution · random cubic-regularization problemsapproximation errorImproved
subspace dimension t much smaller than sqrt of condition numberKrylov subspace solution · random cubic-regularization problemsapproximation error: decays as t^-2Improved
perturbation scheme with tuned sigma instead of joint subspace randomization (ablation)Randomized Krylov method · hard-case instances (kappa=1)convergence speed to a fixed target accuracyImproved
perturbation scheme with fixed sigma instead of joint subspace randomization (ablation)Randomized Krylov method · hard-case instances (kappa=1)final accuracyHurt

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