Model
TimeGAN
A generative adversarial network model designed for synthesizing time series data. It combines supervised and unsupervised learning objectives to generate synthetic data that preserves temporal dependency structures characteristic of the original time series.
Timeline (last 90 days)
Oct 10
The TimeGAN paper adopted discriminative scores and predictive scores (train-on-synthetic, test-on-real) as a downstream task evaluation framework for time series generation.
Unverified50%