The Interplay Between Interpolation and Aggregation in Regression: Optimal Sample Complexity

ORID qXlovWytwg · tags icml2026-repro paper-qXlovWytwg

#StatusPageArtifactClaim excerpt
1VERIFIED 2/201-defines-gamma-graph-dimension-definition-viaartifactThe paper defines the gamma-graph dimension d_γ (Definition 3.2) via a gamma-sha…
2VERIFIED 2/202-any-interpolator-based-aggregation-proper-aggregartifactAny interpolator-based aggregation using proper aggregation rules requires sampl…
3VERIFIED 2/203-there-exist-hypothesis-classes-graphartifactThere exist hypothesis classes with γ-graph dimension d_γ and γ-OIG dimension at…
4VERIFIED 2/204-some-hypothesis-classes-constant-oigartifactFor some hypothesis classes with constant γ-OIG dimension, no finite interpolati…
5VERIFIED 2/205-taking-median-three-independently-trainedartifactTaking the median of three independently trained interpolators achieves expected…
6VERIFIED 2/206-proper-learners-require-sample-complexityartifactProper learners require sample complexity Ω((d_γ/ε)ln(1/ε)), strictly worse than…

Open logbook index · logbook.json