Using class-average private images as anchors and a calibrated exponential mechanism, PCEvolve produces differential-privacy synthetic images from a few examples per class, outperforming Private Evolution on four specialized datasets.
Accessed: 2025- 01-22
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CR 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
PCEvolve: Private Contrastive Evolution for Synthetic Dataset Generation via Few-Shot Private Data and Generative APIs
Using class-average private images as anchors and a calibrated exponential mechanism, PCEvolve produces differential-privacy synthetic images from a few examples per class, outperforming Private Evolution on four specialized datasets.