Symmetries and control in generative neural nets

We study generative nets which can control and modify observations, after being trained on real-life datasets. In order to zoom-in on an object, some spatial, color and other attributes are learned by classifiers and attention nets. Plugging these symmetry statistics in the generative layers of auto-classifiers-encoders (ACE) appears to be the most direct way to simultaneously: i) generate new observations with arbitrary attributes, from a given class , ii) describe the true "style", i.e., the low-dimensional latent manifold encoding the "essence" of the data, after superfluous attributes are factored out, and iii) organically control, i.e., move or modify objects within given observations. We demonstrate the sharp improvement of the generative qualities of shallow ACE with spatial symmetry statistics, on the distorted MNIST and CIFAR10 datasets.