Generative Adversarial Nets from a Density Ratio Estimation Perspective
This work provides a theoretical reinterpretation of GANs that could enhance stability and performance in generative modeling, though it appears incremental as it builds on existing density ratio estimation concepts.
The paper tackles the problem of understanding and improving generative adversarial networks (GANs) by proposing a novel algorithm that repeats density ratio estimation and f-divergence minimization, offering a new perspective that leverages insights from density ratio estimation research.
Generative adversarial networks (GANs) are successful deep generative models. GANs are based on a two-player minimax game. However, the objective function derived in the original motivation is changed to obtain stronger gradients when learning the generator. We propose a novel algorithm that repeats the density ratio estimation and f-divergence minimization. Our algorithm offers a new perspective toward the understanding of GANs and is able to make use of multiple viewpoints obtained in the research of density ratio estimation, e.g. what divergence is stable and relative density ratio is useful.