Data Tech

Much of the data collected by corporations and public institutions is too sensitive to share publicly, or even with third parties. Medical records in particular are very difficult to release to outside researchers. Generative Adversarial Networks (GANs) may provide a solution to this problem: a GAN can be trained to generate new data that is representative of real data, but without confidentiality issues. In this talk, I will provide a general introduction to GANs, and present results showing that successful models can be trained using only generated data.