Time-series Generative Adversarial Networks (TimeGAN) is the work of Jinsung Yoon, Daniel Jarrett, and Mihaela van der Schaar (paper). This repository implements TimeGAN (original code) with TensorFlow 2.X version, mainly for the Hide-and-seek privacy challenge held by NeurIPS (webpage).
cat requirements.txt | xargs -n 1 pip installExecute with default settings (use the Stock dataset)
python main.pyExecute with differential privacy
python main.py --use_dpsgd True