Linear Regression with Unknown Truncation Beyond Gaussian Features

ORID DsV89lJ58l · tags icml2026-repro paper-DsV89lJ58l

#StatusPageArtifactClaim excerpt
1VERIFIED 2/201-gives-first-polynomial-time-poly-algorithmartifactThe paper gives the first polynomial-time, poly(d/ε), algorithm for truncated li…
2VERIFIED 2/202-algorithm-assumes-known-lower-boundartifactThe algorithm assumes a known lower bound α on the truncation/survival probabili…
3VERIFIED 2/203-phase-algorithm-learns-unknown-survivalartifactPhase I of the algorithm learns the unknown survival set as a union of k interva…
4VERIFIED 2/204-phase-estimates-regression-vector-viaartifactPhase II estimates the regression vector via projected stochastic gradient desce…
5VERIFIED 2/205-sub-gaussianity-implies-smoothness-auxiliary-disartifactLemma 3.4 shows sub-Gaussianity implies smoothness of the auxiliary distribution…
6VERIFIED 2/206-improves-prior-work-lmz24a-requiredartifactThis improves on prior work (LMZ24a) which required d^poly(k/ε) running time and…

Open logbook index · logbook.json