Categorical Reparameterization with Gumbel-Softmax

@article{Jang2016CategoricalRW,
  title={Categorical Reparameterization with Gumbel-Softmax},
  author={Eric Jang and Shixiang Gu and Ben Poole},
  journal={CoRR},
  year={2016},
  volume={abs/1611.01144}
}
Categorical variables are a natural choice for representing discrete structure in the world. However, stochastic neural networks rarely use categorical latent variables due to the inability to backpropagate through samples. In this work, we present an efficient gradient estimator that replaces the non-differentiable sample from a categorical distribution with a differentiable sample from a novel Gumbel-Softmax distribution. This distribution has the essential property that it can be smoothly… CONTINUE READING
Highly Influential
This paper has highly influenced 50 other papers. REVIEW HIGHLY INFLUENTIAL CITATIONS
Highly Cited
This paper has 366 citations. REVIEW CITATIONS
253 Extracted Citations
24 Extracted References
Similar Papers

Citing Papers

Publications influenced by this paper.
Showing 1-10 of 253 extracted citations

367 Citations

0100200300201620172018
Citations per Year
Semantic Scholar estimates that this publication has 367 citations based on the available data.

See our FAQ for additional information.

Referenced Papers

Publications referenced by this paper.
Showing 1-10 of 24 references

Similar Papers

Loading similar papers…