Privately generating tabular data using language models
This addresses privacy concerns for data sharing in applications like healthcare or finance, but it is incremental as it adapts existing language model methods to tabular data.
The paper tackled the problem of generating synthetic tabular data with privacy by treating each row as a sentence and training a language model with differential privacy, achieving competitive results across multiple datasets even at small scales.
Privately generating synthetic data from a table is an important brick of a privacy-first world. We propose and investigate a simple approach of treating each row in a table as a sentence and training a language model with differential privacy. We show this approach obtains competitive results in modelling tabular data across multiple datasets, even at small scales that favor alternative methods based on marginal distributions.