sdv-dev/CTGAN

Conditional GAN for generating synthetic tabular data.

What it solves

It provides a way to generate high-fidelity synthetic versions of single-table tabular data. This allows users to create artificial datasets that maintain the statistical properties of real-world data without using the original records.

How it works

The library implements deep learning models, specifically the Conditional GAN (CTGAN) and TVAE (Tabular Variational Autoencoder), based on a 2019 NeurIPS paper. These models learn the distribution of a real dataset and can then be used to sample new, synthetic records that mimic the original data.

Who it’s for

Data scientists and ML engineers who need synthetic tabular data for testing, privacy-preserving data sharing, or augmenting datasets.

Highlights

  • Uses deep learning architectures (CTGAN and TVAE) specifically designed for tabular data.
  • Supports both continuous (float) and discrete (int/string) data types.
  • Can be used as a standalone library or as part of the larger Synthetic Data Vault (SDV) ecosystem.
  • Based on peer-reviewed research presented at NeurIPS 2019.

Related

  • Project
    sdv-dev/SDV

    A Python library for creating tabular synthetic data using machine learning to emulate real-world patterns while maintaining privacy.

  • Project
    hitsz-ids/synthetic-data-generator

    A specialized framework for generating high-quality, privacy-preserving synthetic tabular data using statistical models, GANs, and LLMs.

  • Project
    Data-Centric-AI-Community/fg-data-synthetic

    A package for generating synthetic tabular and time-series data using generative models like GANs and Gaussian Mixtures to ensure privacy and augment datasets.

  • Project
    mostly-ai/mostlyai

    A Python toolkit for creating high-fidelity, privacy-safe synthetic tabular and language data to enable secure data sharing and analysis.

  • Project
    tabularis-ai/be_great

    A framework that uses pretrained Transformer language models to generate high-quality synthetic tabular data, supporting both training-based and schema-based generation.