Matt Hergott - A Leap into the Future: Generative Adversarial Networks (GANs)

Session Title: A Leap into the Future: Generative Adversarial Networks (GANs)

Description:
Generative adversarial networks (GANs), which consist of two neural networks learning from each other in a game, have the potential to help machine learning systems make a quantum leap forward in intelligence. Introduced just a few years ago, these deep learning instruments have already shown promise in a wide variety of futuristic applications.

This presentation will give a high-level overview of GANs, explain the innovations inherent in prominent GAN models, describe some of the many potential uses of GANs, and walk through a learning process to train a GAN to create realistic financial data for simulation analysis.

Some of the GAN architectures we’ll cover include:
• The original GAN
• The Wasserstein GAN with gradient penalty (WGAN-GP)
• CycleGAN
• DiscoGAN
• FusionGAN
• RadialGAN

Speaker Bio:

Matt Hergott (MiaBella AI) has a long history in quantitative finance, and he currently focuses on traditional econometric analysis, deep learning neural networks, and GPU-accelerated computing. He also has expertise in complex quantitative visualizations, and his interactive holographic statistical application achieved Microsoft’s top-16 list of “noteworthy HoloLens content.”