Calibrating Energy-based Generative Adversarial Networks
This addresses a limitation in GANs for researchers and practitioners by providing a method to obtain direct energy estimates, which is incremental as it builds on existing GAN frameworks.
The paper tackles the problem of enabling Generative Adversarial Networks (GANs) to estimate energy or density for samples, proposing a training framework that ensures convergence to the true data distribution and allows the discriminator to retain density information, with empirical results verifying the discriminator recovers the energy of the data distribution.
In this paper, we propose to equip Generative Adversarial Networks with the ability to produce direct energy estimates for samples.Specifically, we propose a flexible adversarial training framework, and prove this framework not only ensures the generator converges to the true data distribution, but also enables the discriminator to retain the density information at the global optimal. We derive the analytic form of the induced solution, and analyze the properties. In order to make the proposed framework trainable in practice, we introduce two effective approximation techniques. Empirically, the experiment results closely match our theoretical analysis, verifying the discriminator is able to recover the energy of data distribution.