Natural Image Manipulation for Autoregressive Models Using Fisher Scores
This addresses a limitation for researchers and practitioners using autoregressive models in image generation, though it is incremental as it builds on existing embedding techniques.
The paper tackled the problem of controlled sample generation in autoregressive models by proposing Fisher scores for extracting embeddings, resulting in more meaningful sample manipulation compared to alternative methods like network activations.
Deep autoregressive models are one of the most powerful models that exist today which achieve state-of-the-art bits per dim. However, they lie at a strict disadvantage when it comes to controlled sample generation compared to latent variable models. Latent variable models such as VAEs and normalizing flows allow meaningful semantic manipulations in latent space, which autoregressive models do not have. In this paper, we propose using Fisher scores as a method to extract embeddings from an autoregressive model to use for interpolation and show that our method provides more meaningful sample manipulation compared to alternate embeddings such as network activations.