REVIEW 2 cited by
GANS for Sequences of Discrete Elements with the Gumbel-softmax Distribution
Not yet reviewed by Pith; the record is open.
This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.
SPECIMEN: schema-true, not a live event
T0 review · schema-true
One-sentence machine reading of the paper's core claim.
pith:XXXXXXXX · record.json · timestamp
read the original abstract
Generative Adversarial Networks (GAN) have limitations when the goal is to generate sequences of discrete elements. The reason for this is that samples from a distribution on discrete objects such as the multinomial are not differentiable with respect to the distribution parameters. This problem can be avoided by using the Gumbel-softmax distribution, which is a continuous approximation to a multinomial distribution parameterized in terms of the softmax function. In this work, we evaluate the performance of GANs based on recurrent neural networks with Gumbel-softmax output distributions in the task of generating sequences of discrete elements.
Forward citations
Cited by 2 Pith papers
-
Joint-stochastic-approximation Autoencoders with Application to Semi-supervised Learning
A stochastic-approximation autoencoder that maximizes the true log-likelihood and uses MCMC-corrected posterior sampling is applied to semi-supervised learning with discrete latent codes, reporting useful but not stat...
-
FlexAct: Why Learn when you can Pick?
A Gumbel-Softmax router that discretely selects among five fixed activation functions, plus a gradient-norm regularizer, recovers the generating activation on toy regression tasks but never beats the matching fixed ac...
Discussion (0). Continue with ORCID to comment.