GANs for Automatic Generation of Data Plots

João Tomás Caldeira, Cláudia Antunes

Abstract: In the last years, online learning saw a surge in relevance, along with interest in the automation of the learning process. Variety and amount of practice exercises are of particular importance, since they limit the amount of practice a student can have. In this paper, we propose the use of generative adversarial networks (GANs) to produce scatter plots, for students’ training on data profiling tasks in data-driven courses. Our results show that progressively grown GANs (ProGANs) generate scatter plots with little tuning and display an adequate level of randomness, diversity, and quality. These properties show promise in allowing for diversity of exercises created from the generated plots.

External IDs:dblp:conf/adma/CaldeiraA22