AI Paper Atlas

Research map / Federated learning and privacy / Differential privacy

Differentially Private Synthetic Data via APIs 4: Tabular Data

Toan Tran, Arturs Backurs, Zinan Lin, Victor Reis, Li Xiong, Sergey Yekhanin

ICML 2026 · Federated learning and privacy › Differential privacy · approach: Differentially private synthetic data · 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
replace XOR-stress-test baselines' marginal queries with Tab-PE full-record nearest-neighbor votingTab-PE · XOR simulation (5 features, eps=1.0)AUC: baselines random guess -> Tab-PE 0.8Improved
use random-walk variation with scheduled mutation decay instead of crossover and mutation operators (ablation)Tab-PE · Artificial Charactersaccuracy: 43% -> 50%Improved
use two-stage selection (sampling then ranking) instead of ranking-only or sampling-only (ablation)Tab-PE · Artificial Characters (eps=1.0)validation accuracy and 5-WDImproved
sampling-only selection (no ranking) in later iterations (ablation)Tab-PE · Artificial Characters (eps=1.0)downstream accuracyHurt
polynomial decay of mutation rate instead of linear decay (ablation)Tab-PE · Artificial CharactersutilityImproved
increase marginal query degree for baselines (ablation)query-based baselines (AIM, GEM, PrivGSD) · NN prior SCM datasetaccuracy and runtimeNo clear effect
generate synthetic samples at 10-20% of the original dataset size (ablation)Tab-PE · eps=1.0downstream utilityImproved
run Tab-PE for 15-20 iterations rather than fewer (ablation)Tab-PE · hyperparameter sensitivityutilityImproved
use Tab-PE instead of AIM on low-order real datasetsTab-PE · widely used low-order datasetsaccuracy: about 1% accuracy drop vs AIMHurt
use Tab-PE instead of AIM on Tree and RFF SCM priorsTab-PE · SCM simulation, eps=1.0accuracy: roughly 10% gains over best baselines; about 4% for NN priorImproved
use Tab-PE instead of AIM at low-order marginal fidelityTab-PE · Artificial Characters and Person Activity1-3 way WDHurt
use Tab-PE instead of AIM for runtimeTab-PE · NN prior SCM, eps=1.0, CPUrunning time: 28x faster than AIM, 10x faster than PrivMRFImproved

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