%0 Journal Article %1 gal2015dropout %A Gal, Yarin %A Ghahramani, Zoubin %D 2015 %K bayesian deep-learning uncertainty %T Dropout as a Bayesian Approximation: Appendix %U http://arxiv.org/abs/1506.02157 %X We show that a neural network with arbitrary depth and non-linearities, with dropout applied before every weight layer, is mathematically equivalent to an approximation to a well known Bayesian model. This interpretation might offer an explanation to some of dropout's key properties, such as its robustness to over-fitting. Our interpretation allows us to reason about uncertainty in deep learning, and allows the introduction of the Bayesian machinery into existing deep learning frameworks in a principled way. This document is an appendix for the main paper "Dropout as a Bayesian Approximation: Representing Model Uncertainty in Deep Learning" by Gal and Ghahramani, 2015.

@article{gal2015dropout, abstract = {We show that a neural network with arbitrary depth and non-linearities, with dropout applied before every weight layer, is mathematically equivalent to an approximation to a well known Bayesian model. This interpretation might offer an explanation to some of dropout's key properties, such as its robustness to over-fitting. Our interpretation allows us to reason about uncertainty in deep learning, and allows the introduction of the Bayesian machinery into existing deep learning frameworks in a principled way. This document is an appendix for the main paper "Dropout as a Bayesian Approximation: Representing Model Uncertainty in Deep Learning" by Gal and Ghahramani, 2015.}, added-at = {2019-09-09T14:30:56.000+0200}, author = {Gal, Yarin and Ghahramani, Zoubin}, biburl = {https://www.bibsonomy.org/bibtex/2536ace6c9f0dc30becbf4d87a7f99a20/kirk86}, description = {[1506.02157] Dropout as a Bayesian Approximation: Appendix}, interhash = {8a82d52d5505d8ef3f86a44687826feb}, intrahash = {536ace6c9f0dc30becbf4d87a7f99a20}, keywords = {bayesian deep-learning uncertainty}, note = {cite arxiv:1506.02157Comment: 20 pages, 1 figure; ICML proceedings version}, timestamp = {2019-09-09T14:30:56.000+0200}, title = {Dropout as a Bayesian Approximation: Appendix}, url = {http://arxiv.org/abs/1506.02157}, year = 2015 }