Development of Data Augmentation Techniques for Timeseries Data (Masterarbeit)

Kurzfassung

This study explores the application of advanced generative models to increase the volume of dataset for material testing. The challenges of working with limited data are addressed through data augmentation and synthetic data generation techniques. We investigate the performance of Generative Adversarial Networks (GANs), Variational Autoencoders (VAEs), and their variants, with a focus on hyperparameter optimization. The capabilities of those models are evaluated using Frechet Inception Distance (FID) as metric. Results indicate that Wasserstein GAN with Gradient Penalty (WGAN-GP) and its variants outperform traditional GANs and VAEs in producing realistic data.