gan

Proposes TriHead-GAN, a transformer-based GAN with a triple-head discriminator that jointly supervises distributional authenticity, cross-variable dependencies, and temporal smoothness for generating realistic carbon emission time series, outperforming baselines on multiple datasets.

This paper introduces SwiGAN, a Conditional GAN-based framework for generating future spatio-temporal trajectories of climatic indices to assist in drought risk management and insurance strategy.