Generator Reversal
This addresses the limitation of simple priors in generative models for machine learning, offering improved control and performance, though it appears incremental as it builds on existing generator-based paradigms.
The paper tackles the problem of training generative models by proposing to use more flexible code distributions estimated non-parametrically through generator reversal, resulting in benefits such as more powerful models and better latent structure modeling.
We consider the problem of training generative models with deep neural networks as generators, i.e. to map latent codes to data points. Whereas the dominant paradigm combines simple priors over codes with complex deterministic models, we propose instead to use more flexible code distributions. These distributions are estimated non-parametrically by reversing the generator map during training. The benefits include: more powerful generative models, better modeling of latent structure and explicit control of the degree of generalization.