Adaptive Computation Time for Recurrent Neural Networks

arXiv: Neural and Evolutionary Computing, Volume abs/1603.08983, 2016.

Cited by: 82|Bibtex|Views119
EI
Other Links: dblp.uni-trier.de|academic.microsoft.com|arxiv.org

Abstract:

This paper introduces Adaptive Computation Time (ACT), an algorithm that allows recurrent neural networks to learn how many computational steps to take between receiving an input and emitting an output. ACT requires minimal changes to the network architecture, is deterministic and differentiable, and does not add any noise to the paramete...More

Code:

Data:

Full Text
Your rating :
0

 

Tags
Comments