generative modeling

[A] rhizome is not amenable to any structural or generative model. It is a stranger to any idea of genetic axis or deep structure. A genetic axis is like an objective pivotal unity upon which successive stages are organised; a deep structure is more like a base sequence that can be broken down into another, transformational and subjective, dimension. This does not constitute a departure from the representative model of the tree or root - pivotal taproot or fascicles (for example, Chomsky’s “tree” is associated with a base sequence and represents the process of its own generation in terms of binary logic). A variation on the oldest form of thought. It is our view that genetic axis and profound structure are above all infinitely reproducible principles of tracing. All of tree logic is a logic of tracing and reproduction. In linguistics as in psychoanalysis, its object is an unconscious that is itself representative, crystallised into codified complexes, laid out along a genetic axis and distributed within a syntagmatic structure. Its goal is to describe a de facto state, to maintain balance in intersubjective relations, or to explore an unconscious that is already there from the start, lurking in the dark recesses of memory and language.
—  A Thousand Plateaus - Gilles Deleuze & Félix Guattari
4

Neural Photo Editing with Introspective Adversarial Networks

Proof of concept application by Andrew Brock uses neural network trained data to modify features in portrait photographs:

We present the Neural Photo Editor, an interface for exploring the latent space of generative image models and making large, semantically coherent changes to existing images. Our interface is powered by the Introspective Adversarial Network, a hybridization of the Generative Adversarial Network and the Variational Autoencoder designed for use in the editor. Our model makes use of a novel computational block based on dilated convolutions, and Orthogonal Regularization, a novel weight regularization method. We validate our model on CelebA, SVHN, and ImageNet, and produce samples and reconstructions with high visual fidelity. 

More Here