AI Paper Atlas

Research map / Federated learning and privacy / Differential privacy

Differentially Private and Fair Classification via Calibrated Functional Mechanism

Jiahao Ding, Xinyue Zhang, Xiaohuan Li, Junyi Wang, Rong Yu, Miao Pan

AAAI 2020 · Federated learning and privacy › Differential privacy · approach: Private training with public data · 50 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
Use attribute-calibrated noise in place of PFLR*'s approachPDFC and ADFC · USrisk difference: 0.037 ± 0.038 -> 0.003 ± 0.034 (PDFC); 0.037 ± 0.038 -> 0.004 ± 0.007 (ADFC)Improved
Use attribute-calibrated noise in place of PFLR*'s approachPDFC and ADFC · Adultrisk difference: 0.045 ± 0.095 -> 0.048 ± 0.108 (PDFC); 0.045 ± 0.095 -> 0.146 ± 0.131 (ADFC)Mixed
Use ADFC instead of PFLR*ADFC · Adult accuracy across privacy budgetsaccuracyImproved
Use PDFC instead of PFLR*PDFC · Adult accuracy across privacy budgetsaccuracyMixed
Replace Laplace noise with Gaussian noise using the relaxed functional mechanismRelaxed functional mechanism · US accuracy across privacy budgetsaccuracyImproved
Decrease δ, increasing Gaussian noise scaleRelaxed functional mechanism · US accuracyaccuracyHurt
Choose different unprotected attributes for x_sPDFC and ADFC · Adult accuracyaccuracyMixed
Use attribute-calibrated noise rather than PFLR*PDFC and ADFC · Adult and US, risk difference across εrisk differenceMixed

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