Adaptação do WGAN ao processo estocástico

Abstract

Within different areas of knowledge, data (different types of information) are valuable and their analysis is even more valuable. Then, associating the area of artificial intelligence, a new trend is observed, the generation of synthetic data to fill the lack of data. Therefore, analyzing current contexts, this work aims to demonstrate the use of techniques based on random events to optimize the result in the execution of algorithms based on GAN (Generative adversarial networks) and through a validation through the calculating the FID (Frechet Inception Distance) it was possible to analyze the results, determining the quality of the data generated by the proposed algorithm compared to WGAN.