Войти
аналитика - Privacy-Preserving GAN Training
Training Generative Models from Privatized Data via Entropic Optimal Transport
The authors propose a framework for training Generative Adversarial Networks (GANs) on differentially privatized data using entropic optimal transport, enabling the generator to learn the raw data distribution even with access to privatized samples.
Training generative models from privatized data using Entropic Optimal Transport
Entropic regularization of optimal transport enables learning from privatized data while mitigating noise effects and curse of dimensionality.
Продукты | Ресурсы
© 2024 by Linnk AI