Synthetic Data Generators -- Sequential and Private

Authors: Olivier Bousquet, Roi Livni, Shay Moran

NeurIPS 2020 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Theoretical We study the sample complexity of private synthetic data generation over an unbounded sized class of statistical queries, and show that any class that is privately proper PAC learnable admits a private synthetic data generator (perhaps nonefficient). Previous work on synthetic data generators focused on the case that the query class D is finite and obtained sample complexity bounds that scale logarithmically with the size |D|. Here we construct a private synthetic data generator whose sample complexity is independent of the domain size, and we replace finiteness with the assumption that D is privately PAC learnable (a formally weaker task, hence we obtain equivalence between the two tasks).
Researcher Affiliation Collaboration Olivier Bousquet Google Research, Brain team Zürich obousquet@google.com Roi Livni Tel-Aviv University Tel-Aviv, Israel rlivni@tauex.tau.ac.il Shay Moran Technion Haifa, Israel shaymoran1@gmail.com
Pseudocode No The paper does not contain structured pseudocode or algorithm blocks.
Open Source Code No The paper does not provide any statement or link regarding the availability of open-source code for the methodology described.
Open Datasets No This is a theoretical paper that focuses on mathematical proofs and characterizations, and therefore does not use publicly available datasets for training empirical models.
Dataset Splits No This is a theoretical paper and does not conduct experiments with data, thus it does not provide dataset splits for training, validation, or testing.
Hardware Specification No The paper does not provide any specific hardware details as it is a theoretical work and does not involve empirical experiments.
Software Dependencies No The paper does not mention any specific software dependencies with version numbers, as it is a theoretical work.
Experiment Setup No The paper is theoretical and does not describe an experimental setup with hyperparameter values or training configurations.