We aim to generate realistic images from text descriptions using GAN architecture. The network that we have designed is used for image generation for two datasets: MSCOCO and CUBS.

A kind of neuralnet that runs in browser where each node smoothly chooses between many neural activation functions (sine tanh log exp + * arcsine etc) and is trained without backprop, instead using calculus directly on the sum of squared loss of all weights and all training data at once. Not GPU optimized yet.

Differentiable AURC