Transfer learning for time series classification using synthetic data generation

Both time series classification and transfer learning have increasingly been the focus of research in recent years. However, only a limited number of studies have combined time series classification with transfer learning. Time series classification (TSC) is the task of training a classifier to map a given time series input to a probability distribution over the possible class values. Typically, transfer learning (TL) algorithms learn from a source dataset and task and then apply the knowledge gained to another target dataset and task. TL has received considerable attention in the domains of computer vision and natural language processing, but less research attention has been devoted to the task of TSC, which is lacking a state-of-the-art pretrained model that can serve as a good starting point for new TSC tasks.