AI Paper Atlas

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

Recovery Guarantee of Non-negative Matrix Factorization via Alternating Updates

Yuanzhi Li, Yingyu Liang, Andrej Risteski

NeurIPS 2016 · Learning theory and algorithms › Matrix and tensor methods · approach: Matrix completion and factorization · 16 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
Add ReLU thresholding (offset alpha) to the decoding step of the alternating updatealternating decode-update NMF algorithm (purification) · theoretical analysis, simplified sparse weights with adversarial noiserecovery error of feature matrixImproved
Use unbiased zero-mean noise instead of adversarial noise (N2 vs N1)alternating decode-update NMF algorithm · theoretical analysis, simplified casemaximum tolerable noise level relative to adversarial case: noise level sqrt(n) larger than adversarial caseImproved
Use minimum-infinity-norm pseudo-inverse for decoding rather than an arbitrary pseudo-inversedecoding step · theoretical analysisrobustness to noiseImproved
Pre-processing step that rebalances feature scales (D scaling) to relax the xsu <= 2 xsl proportion restrictiongeneral-proportion setting · theoretical analysisapplicability of guaranteeNo clear effect
Use max_i row-wise potential instead of coupled potential a+beta*b for analysisconvergence analysis · theoretical analysispotential function monotonicityNo 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