Gradient Estimation with Stochastic Softmax Tricks
By Max B. Paulus, Dami Choi, Daniel Tarlow, Andreas Krause, Chris J. Maddison. 2020
101 citations recorded by the source corpus.
The Gumbel-Max trick is the basis of many relaxed gradient estimators. These estimators are easy to implement and low variance, but the goal of scaling them comprehensively to large combinatorial distributions is still outstanding. Working within the perturbation model framework, we introduce stochastic softmax tricks, which generalize the Gumbel-Softmax trick to combinatorial spaces. Our framework is a unified perspective on existing relaxed estimators for perturbation models, and it contains many novel relaxations. We design structured relaxations for subset selection, spanning trees, arborescences, and others. When compared to less structured baselines, we find that stochastic softmax tricks can be used to train latent variable models that perform better and discover more latent structure.
Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition, Optimization & Training, Probabilistic & Bayesian Methods
How researchers see this paper
The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.
In Artificial Intelligence
- Richard I. Hartley
- Dylan Campbell
- Stephen Gould
- Devi Parikh
- Zhe Gan
- L. V. D. van der Maaten
- David G. Lowe
- Marius Muja
In Computer Vision and Pattern Recognition
- Dylan Campbell
- Stephen Gould
- Richard I. Hartley
- Luc Van Gool
- Xiaogang Wang
- Dacheng Tao
- Cordelia Schmid
- Trevor Darrell
In Optimization & Training
In Probabilistic & Bayesian Methods
Citations in the corpus
- DSelect-k: Differentiable Selection in the Mixture of Experts with Applications to Multi-Task Learning (2021, neutral): “A related line of work focuses on stochastic k-subset selection in neural networks, e.g., see [26, 5, 39] and the references therein.”
- Deep Declarative Networks (2019, background): “Here the derivative is not well defined and some recent works look at stochastic smoothing to get around this difficulty [8], [45].”