Painting with Fire: A History in GANs: Training Fire

Progressing over years the GAN resolution grows learning training all is unclear. Not quite realistic new methods are found to attain photorealism, but what of the strange strange land of the GAN. Channeling the fire myths of the world, Prometheus, the Phoenix, the gods of the sun and flame. A story of technology brought to humans; to myths of destruction and rebirth. To be immortal; to be a god; it is always a paradox of fire and technology. The morality of the fire is always in question but we desire it and fear it. And yet it has changed us, we have learned so much since we first touched fire. We used it to destroy the wood, smelt the metal and build the chip. And now we teach and show the fire what we know; what we create and our reflection is there telling us our faults. The GAN is but a single neuron in a blazing star; a greater fire is coming. One of 200 unique pieces of art in Painting With Fire: A History in GANs. Exploring the history of generative adversarial networks GAN as represented by fire. https://bardionson.com/painting-with-fire DCGAN - Radford, Alec; Metz, Luke; Chintala, Soumith . "Unsupervised Representation Learning with Deep Convolutional Generative Adversarial Networks" Wasserstein GAN (WGAN): Introduced by Martin Arjovsky, Soumith Chintala, Léon Button in 2017, WGAN addressed training stability issues in GANs by using a different approach to measure the distance between generated and real data distributions. This innovation led to more stable training of GANs and improved quality of generated images.|Jan 2017|https://arxiv.org/abs/1701.07875