AI Paper Atlas

Research map / Learning theory and algorithms / Sample complexity

When are Local Queries Useful for Robust Learning?

Pascale Gourdeau, Varun Kanade, Marta Kwiatkowska, James Worrell

NeurIPS 2022 · Learning theory and algorithms › Sample complexity · approach: Robust and surrogate-loss learning bounds · 0 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 ρ-local equivalence queries to random examplesDistribution-free exact-in-the-ball robust learning · Concept classes with robustly consistent learnersrobust learnabilityImproved
Use local equivalence queries with radius λ strictly smaller than ρMonotone conjunctions and superclasses · Distribution-free exact-in-the-ball robust learningexpected robust risk: \frac{1}{2}Hurt
Replace local membership queries with local equivalence queriesConjunctions · Distribution-free exact-in-the-ball robust learningquery complexity: O\left( \frac{1}{\epsilon}\left(n+\log\frac{1}{\delta}\right)\right) versus 2^{\rho-1}Improved
Set the local equivalence query radius equal to the adversary's budgetMonotone conjunctions · Distribution-free robust learningrobust learnabilityImproved
Set the local equivalence query radius below the adversary's budgetMonotone conjunctions and classes containing them · Distribution-free robust learningrobust learnabilityHurt
Restrict the adversary to precision-bounded counterexamplesHalfspaces in ℝ^n · Distribution-free ρ-robust learningquery complexity: m(n,\epsilon,\delta)\cdot\frac{B^2}{\tau^2}Improved
Assume a margin only on the data distribution's supportHalfspaces in ℝ^n · Robust learning with ρ-LEQquery complexityHurt
Require the margin condition throughout the ρ-expansion of the distribution supportHalfspaces in ℝ^n · Robust learning with ρ-LEQquery complexity: r=\frac{mB^2}{\gamma^2}Improved

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