Adversarial Learning of a Sampler Based on an Unnormalized Distribution
This addresses a challenge in generative modeling for scenarios where sampling from the target distribution is difficult, though it appears incremental as it builds on existing GAN frameworks.
The paper tackles the problem of adversarial learning when only an unnormalized density is available, extending GANs to this setting and developing new regularization concepts, with encouraging results demonstrated in applications like deep soft Q-learning.
We investigate adversarial learning in the case when only an unnormalized form of the density can be accessed, rather than samples. With insights so garnered, adversarial learning is extended to the case for which one has access to an unnormalized form u(x) of the target density function, but no samples. Further, new concepts in GAN regularization are developed, based on learning from samples or from u(x). The proposed method is compared to alternative approaches, with encouraging results demonstrated across a range of applications, including deep soft Q-learning.