Papers of the day   All papers

Gradient Estimation with Stochastic Softmax Tricks

Comments

Chris J. Maddison: Stochastic Softmax Tricks: We generalize the Gumbel-Softmax, and introduce new tricks for backpropping through all kinds of random discrete objects: spanning trees, subset selection, & more (https://arxiv.org/abs/2006.08063). First authors @mbpaulus, @damichoi95. https://t.co/45xaHtVsx8

5 replies, 1055 likes


will grathwohl: HOT SHIT ALERT

2 replies, 44 likes


Thomas Kipf: With Stochastic Softmax Tricks (https://arxiv.org/abs/2006.08063), Neural Relational Inference can be used to discover other forms of latent structure, such as spanning trees. Figure from the Stochastic Softmax Tricks paper. https://t.co/ArgQkuGVL2

0 replies, 35 likes


Dr Simon Osindero: Very cool! Further expanding the set of backprop-able building blocks for model architectures.

1 replies, 9 likes


HotComputerScience: Most popular computer science paper of the day: "Gradient Estimation with Stochastic Softmax Tricks" https://hotcomputerscience.com/paper/gradient-estimation-with-stochastic-softmax-tricks https://twitter.com/cjmaddison/status/1272899088380502016

0 replies, 2 likes


arXiv in review: #NeurIPS2020 Gradient Estimation with Stochastic Softmax Tricks. (arXiv:2006.08063v1 [stat\.ML]) http://arxiv.org/abs/2006.08063

0 replies, 1 likes


Content

Found on Jun 16 2020 at https://arxiv.org/pdf/2006.08063.pdf

PDF content of a computer science paper: Gradient Estimation with Stochastic Softmax Tricks