Evaluating Lossy Compression Rates of Deep Generative Models
ICML, pp. 4444-4454, 2020.
We found that different Generative Adversarial Networks variants with the same code size have nearly identical rate distortion curves, and that the code size dominates the algorithmic differences of GANs
The field of deep generative modeling has succeeded in producing astonishingly realistic-seeming images and audio, but quantitative evaluation remains a challenge. Log-likelihood is an appealing metric due to its grounding in statistics and information theory, but it can be challenging to estimate for implicit generative models, and sca...More
PPT (Upload PPT)