Time Series Data Augmentation for Deep Learning: A Survey

https://arxiv.org/pdf/2002.12478.pdf

 

Current DGMs(Deep Generative Models) adopted for time series data augmentation are mainly GANs.

In summary, besides the common GAN architectures, how to leverage other deep generative models like DARNs(Deep Autoregressive Networks), NFs(Normalizing Flows), and VAEs(Variational Autoencoders), which are less investigated for time series data augmentation, remain exciting future opportunities.